Bäßler2022_Article_UnsupervisedAnomalyDetectionIn.pdf

Vol.:(0123456789)
Machine Learning (2022) 111:1377–1408
https://doi.org/10.1007/s10994-022-06129-4
1 3
Unsupervised anomaly detection in multivariate time series
with online evolving spiking neural networks
Dennis Bäßler1
· Tobias Kortus1
· Gabriele Gühring1

Received: 20 December 2020 / Revised: 17 December 2021 / Accepted: 24 December 2021 /
Published online: 10 March 2022
© The Author(s) 2022
Abstract
With the increasing demand for digital products, processes and services the research area
of automatic detection of signal outliers in streaming data has gained a lot of attention. The
range of possible applications for this kind of algorithms is versatile and ranges from the
monitoring of digital machinery and predictive maintenance up to applications in analyz-
ing big data healthcare sensor data. In this paper we present a method for detecting anoma-
lies in streaming multivariate times series by using an adapted evolving Spiking Neural
Network. As the main components of this work we contribute (1) an alternative rank-order-
based learning algorithm which uses the precise times of the incoming spikes for adjusting
the synaptic weights, (2) an adapted, realtime-capable and efficient encoding technique for
multivariate data based on multi-dimensional Gaussian Receptive Fields and (3) a continu-
ous outlier scoring function for an improved interpretability of the classifications. Spiking
neural networks are extremely efficient when it comes to process time dependent informa-
tion. We demonstrate the effectiveness of our model on a synthetic dataset based on the
Numenta Anomaly Benchmark with various anomaly types. We compare our algorithm to
other streaming anomaly detecting algorithms and can prove that our algorithm performs
better in detecting anomalies while demanding less computational resources for processing
high dimensional data.
Keywords Anomaly detection · Outlier detection · Evolving Spiking Neural Networks ·
Online learning · Streaming multivariate time series · Machine learning
Editor: Joao Gama.
* Gabriele Gühring
gabriele.guehring@hs-esslingen.de
Dennis Bäßler
debait01@hs-esslingen.de
Tobias Kortus
tokoit02@hs-esslingen.de
1
Hochschule Esslingen, Flandernstr. 101, 73732 Esslingen, Baden‑Württemberg, Germany

1378
1 3
1 Introduction
Detecting outliers in time series data, especially streaming data, has gained significant rel-
evance due to the recent exponential growth in the amount of data captured in big-data and
IoT applications (Ahmad et al. 2017; Munir et al. 2019; Macia̧g et al. 2021). Particularly
the detection of anomalies in streaming time series data places high demands on the devel-
opment of effective and efficient algorithms (Ahmad et al. 2017). Real-time capable algo-
rithms are required for many applications in order to process these generated or recorded
data. In many cases it is not easy to collect sufficient training data with marked anomalies
for the supervised training of an anomaly detector to identify anomalies in data streams
(Macia̧g et al. 2021). It is therefore particularly important to design anomaly detectors that
can correctly classify anomalies from data where none of the input values need to be clas-
sified. Ideally, the designed anomaly detector should learn in an online mode in which the
current input values adjust the parameters of the detector for better anomaly detection of
future input data. Since conventional machine learning algorithms are in many cases una-
ble to cope with these requirements or can only handle them with a large expenditure of
resources, there is a great interest in new efficient solutions. Spiking neural networks have
the capability for processing information in a fast way, both in terms of energy and data,
due to their functional similarity to the brain (Bing et al. 2018a; Gerstner and Kistler 2002;
Lobo et al. 2018) thus they are particularly well suited for online detection of anomalies.
Although they are one of the major exponents of the third generation of artificial neural
networks, it took a while till they were applied to an online learning approach (Lobo et al.
2018, 2020).
In this paper, we extend an existing algorithm for anomaly detection by Maciąg et al.
(2021) from the field of evolving Spiking Neural Networks, in which learning processes,
neural communication and classification of data instances are based exclusively on spike
exchange between neurons. The approach by Maciąg et al. (2021) is however solely
designed for univariate time series. Since multivariate data streams are in practice quiet
common we contribute in this work several appropriate measures to make the algorithm
work for multivariate data. Here, we focused both on the runtime and real-time capability
of the algorithm as well as the general performance for the detection of different types of
anomalies which are presented in appendix A.1. Our results can be summarized in the fol-
lowing way:
– We present an online capable encoding technique for multivariate times series data. Our
method is based on the work of Panuku and Sekhar (2008). It provides a significant per-
formance boost to a parallel execution of several univariate versions of the algorithm of
Maciąg et al. (2021).
– We use a new continuous outlier score, which is adapted from Ahmad et al. (2017).
This outlier score can be executed in an online manner and provides significantly better
results as the currently used technique in Maciąg et al. (2021).
– We apply the SpikeTemp learning approach by Wang et al. (2017) as an improvement
for the existing technique and therefore getting better and faster results for anomaly
detection even in the univariate case.
Compared to deep learning models the main advantage of our model is, that it needs no
time for training and less memory space, while its performance is in a similar range (see
Geiger et al. 2020 and Sect. 5). In the following we use outlier for anomalous data points as

1379
1 3
well as anomaly. The difference between an outlier and an anomaly is small (see Aggarwal
2013) and some authors use both expressions as well, i.e. Shukla and Sengupta (2020).
2 Related work
Due to the extensive demand for efficient and versatile algorithms for outlier detection a
wide variety of approaches has been studied in the existing literature. Depending on the
application domain, different kinds of algorithms and techniques have been used. In the fol-
lowing section we present a general review of currently available related literature sources
regarding to either univariate or multivariate unsupervised and semi-supervised outlier
detection in time series given the applicability of these techniques in a streaming scenario.
One of the most commonly used technique for detecting outliers in a wide variety of
different application domains is the Local Outlier Factor (LOF) as proposed by Breunig
et al. (2000). This technique uses a density measure which describes how isolated a given
instance is with regards to its neighborhood. Pokrajac et al. (2007) proposed an incremen-
tal version of this algorithm that makes it suitable for an efficient detection of outliers in
data streams. Both the static as well as the incremental LOF algorithm are not designed for
high dimensional data as well as for temporal relationships in the input data (Gühring et al.
2019; Pokrajac et al. 2007), thus making it unapplicable for those specific applications.
Statistical autoregressive (AR) models such as ARIMA, VAR, VARIMA have also been
extensively studied in the past for detecting outliers in both univariate and multivariate
time series (Hau and Tong 1989; Moayedi and Masnadi-Shirazi 2008; Li et al. 2019b).
These kinds of models particularly excel by their simplicity and the lack of an initially
required training phase and are therefore also suitable for streaming data processing. How-
ever, since ARIMA models can barely handle nonlinear relationships, their approach to
complex real-world problems is not always satisfactory (Zhang 2003).
Recent advances in deep learning like natural language processing (NLP) or speech rec-
ognition have caused an increasing utilization of various network architectures in outlier
detection research (Chalapathy and Chawla 2019). These range from Convolutional Neural
Networks (CNN), Long Short Term Memory (LSTM) Networks, Autoencoders up to Gen-
erative Adversarial Networks. Both discriminative and generative models are used in this
context. However, due to the pronounced class imbalance and the associated lack of appro-
priate data labels, discriminative models are used far less often. Fu et al. (2019) proposed
a CNN based denoising autoencoder architecture for detecting outliers in time series data.
The autoencoder structure is used for unsupervised feature learning of variable groups
and is complemented by a classification layer that distinguishes normal from abnormal
samples. Another time series anomaly detection method based on deep learning was pro-
posed by Mahotra et al. (2016). With their EncDec-AD LSTM architecture they propose
an encoder decoder scheme for reconstructing the given input signal using the compressed
vector representation of the decoder. In order to compute the likelihood of an anomaly
they model a normal distribution over the determined error vectors. Generative Adversarial
Networks are also extensively present in current research. Geiger et al. (2020) and Li et al.
(2019a) both use an adversarial minimax training approach on an LSTM architecture in
order to detect outliers in time series. Both use the discriminative and the reconstructive
components of the architecture in order to distinguish fake from real data using a multi
component outlier scoring system. Deep learning based architectures, however, require for

1380
1 3
training and inference a vast amount of computing power, making an application of these
approaches in a streaming context especially on low power devices infeasible.
To exploit the advances from both, the traditional approaches for outlier detection and
deep neural networks there is currently an increased use of hybrid models that are either
used in order to improve the quality, the speed or the scalability of the algorithm. Munir
et al. (2019) uses a combination of an ARIMA model and a two layered Convolutional
Neural Network in a residual learning scheme for the prediction of a given time series
where both models complement each other for optimal results. This information is then
further utilized for detecting existing outliers in the signal using the Euclidean distance
between the real and predicted time step. Another hybrid model is proposed by Shukla and
Sengupta (2020). They use a combination of hierarchical clustering and a Long Short Term
Memory Network (LSTM) where the hierarchical clustering condenses similar correlat-
ing input data which are then fed into a LSTM network enabling the architecture to scale
well for high dimensional data. Papadimitriou et al. (2005) proposed a neural network type
architecture (SPRIT) for detecting outliers in data streams by reconstructing a given time
series using a dynamic repository of incrementally estimated principal components which
are approximated by the weights of a hidden neuron. The number of required eigencom-
ponents is determined by a heuristic, which compares the energy of the input data and the
estimated eigencomponents. Since each additional added principal component can sum-
marize further aspects of the signal which cannot be explained by the previous principal
components, an increase of the repository indicates the existence of an outlier in the cur-
rent time step.
Advances in neuroscientific disciplines produced a large number of novel approaches of
biological plausible learning algorithms for the detection of anomalies in signals. A heav-
ily biologically inspired learning approach used in current anomaly detection research are
Spiking Neural Networks (SNN). Xing et al. (2019) proposed a hybrid learning approach
using a combination of an evolving Spiking Neural Network architecture as well as
a restricted Boltzmann machine for detecting outliers in time series. The concept of an
evolving Spiking Neural Network architecture for the efficient processing of data streams
is used in several other publications. Demertzis et al. (2017, 2019) use in multiple papers
an evolving Spiking Neural Network approach based on a simplified Leaky Integrate and
Fire model (LIF) and a Rank Order Coding scheme in a semi-supervised learning scenario.
Maciąg et al. (2021) use a similar architecture for an Online evolving Spiking Neural net-
work to detect signal outliers in time series data in an online mode using the error between
the reconstructed and the given time series. The evolving architecture used by the previ-
ously mentioned approaches enable a continuous adaption to changing statistics of input
data, making them well suited for online learning (Wysoski et al. 2006).
3
Pattern discovery in multivariate time series
In this paper we present an online capable algorithm, which is able to detect anomalies
in multivariate time series. The detection of anomalies in multivariate times series is not
the same as detecting anomalies in a single valued time series, since also correlations
between the different dimensions have to be taken into account. Since this is also one of
the main differences between our paper and the work of Macia̧g et al. (2021), we here
present as a benchmark algorithm the SPIRIT algorithm, another algorithm focusing on
outlier detection of streaming multivariate time series. Streaming Pattern dIscoveRy in

1381
1 3
multIple Time-series (SPRIT) by Papadimitriou et al. (2005) is a fast online capable mul-
tivariate time series algorithm. It is according to Aggarwal (2013) one of the most well
known unsupervised algorithms which is designed to work not only on single data streams
but takes into account a global direction of correlation between different simultaneous data
streams. In order to evaluate multivariate time series for anomaly detection most of the
time a method for dimensionality reduction is used. The SPIRIT algorithm uses an online
principal component analysis for this task. We explain it here in more detail.
The goal of the SPIRIT algorithm is to approximate a representation of the input signal
with the smallest possible number of principal components. The algorithm does not require
an initial training phase as the principal components are continuously updated with incom-
ing data points 𝐱t =
[
xt,1, … , xt,n
]T
of the n-dimensional data stream.
The principal components matrix of the form W ∈ ℝn×k where n is the number of
dimensions of the incoming signal and k is the number of the current principal components
are initially represented by the conical unit vector 𝐮 =
[
u1 … uk
]

. For each time step the
eigencomponents making up the matrix W are updated. Therefore the current data point 𝐱t
is used to initialize the temporary variable ́
𝐱1 which is used to calculate the projection 𝐲 of
́
𝐱i on the eigencomponents of Wi according to yi = WT
i
́
𝐱i
. On the basis of this projection,
the energy di ← 𝜆di + y2
i
is updated and the reconstruction error 𝐞i = ́
𝐱i − yiWi of the ith
projection is calculated. Papadimitriou et al. (2005) introduce in this context the exponen-
tial forgetting factor 𝜆 as a hyperparameter for the model. This makes it possible to control
the influence of time-dependent trends. Values between 0.96 and 0.98 are suggested in
Papadimitriou et al. (2005). Based on the reconstruction error 𝐞i
, the eigencomponents are
optimized using gradient descent with a learning rate of 𝛾 = 1
di

. The previously mentioned
steps are iteratively repeated for i ≤ k for the next eigencomponents with a new
́
𝐱i+1 = ́
𝐱i − yiWi to reduce the remaining reconstruction error.
To estimate the required number of principal components, two additional metrics are
used:
– The continuous average energy Et on the input data at time t.
– The average total energy of the eigencomponents ̃
E(k)
, which is calculated from the sum
of the individual energies of the eigencomponents ̃
Ei,t at time t, obtained from Eq. (2).
Two further hyperparameters, an upper energy limit FE and a lower energy limit fE are
used to determine whether the existing number of principal components k is sufficient or
surplus. If ̃
E(k) fEEt
, a further principal component is required to sufficiently represent
the data i.e. k ← k + 1 . However, if ̃
E(k) FEEt applies, k − 1principal components are suf-
ficient to represent the data with appropriate quality, i.e. k ← k − 1
. Based on the previous
result, either the k-th eigencomponent is removed from the matrix W or, if the increased
(1)
Et =
𝜆(t − 1)Et−1 + ‖𝐱t‖2
t
(2)
̃
Ei,t =
𝜆(t − 1) ̃
Ei,t−1 + y2
i,t
t
(3)
̃
E(k) =
k
∑
i=1
̃
Ei,t

1382
1 3
number of eigencomponents k is smaller than the dimension of the data, a new eigencom-
ponent in the form of the canonical unit vector uk+1 is added to the matrix W.
According to Papadimitriou et al. (2005), the number of eigencomponents k or the
reconstruction error e can be used as a metric to determine the affiliation of the data point
to the class of outliers. In this paper we use the SPIRIT algorithm as a comparison model
to benchmark the performance of our model. In order to provide comparable values we
extend the outlier detection from Papadimitriou et al. (2005) by the same continuous out-
lier score that we use for our anomaly detection model (see Sect. 4.5).
4
Spiking Neural Networks for outlier detection
4.1 Spiking neural networks
The notion of Spiking Neural Networks is used to describe a group of artificial neural net-
works that have their origins in the field of neuroscience. These neuron models, also known
as the third generation of neural networks (Maas 1997), provide a plausible model of the
neuroscientific processes of information processing in biological neurons via the coding
of sensory information in spikes. They encode information from preceding neuron layers
by generating time-varying pulses caused by the counteraction of excitatory and inhibitory
potentials of preceding layers (Gerstner and Kistler 2002). The timing of a spike is for-
mally determined by the positive crossing of a threshold value 𝜗 of the potential u, caused
by an excitatory signal of a preceding neuron (Gerstner and Kistler 2002).
For the mathematical approximation of these complex process, different models exist. But
usually only the Integrate and Fire model of Louis Lapicque (1907) and its variations (Ger-
stner and Kistler 2002) in conjunction with the Rank Order Coding (ROC) model accord-
ing to Thorpe et al. (1998) are of high relevance in technical applications, especially with
regard to the low complexity and the potential to process large amounts of data.
This new type of artificial, biologically inspired networks is characterized, especially
with respect to their structural differences, by their high performance compared to con-
ventional neural networks (Maas 1997). Due to the high potential of Spiking Neural Net-
works, they can be found in research projects of different disciplines such as signal pro-
cessing (Amirshahi and Hashemi 2019), speech recognition (Wu et al. 2020) or control
engineering (Bing et al. 2018b).
4.2 Online evolving Spiking Neural Networks
Evolving Spiking Neural Networks (eSNN) (Wysoski et al. 2008) and their variation of
Online evolving Spiking Neural Networks (OeSNN) represent a subcategory of Spik-
ing Neural Networks. The neural activities are commonly modeled by a simplified
Leaky Integrate and Fire (LIF) neuron model (Schliebs and Kasabov 2013; Wysoski
et al. 2008). Whereas the topology of the adaptive evolving layer changes continuously
with new incoming data from the previous input layer (Watts 2009). In this paper we
use a OeSNN model for anomaly detection developed by Maciąg et al. (2021) as base-
line architecture, which consists exclusively of an input and an output layer without any
(4)
t(f)
∶ u(t(f)
) = 𝜗 and
du(t)
dt
|
|
|
|t=t(f)
0

1383
1 3
hidden layers. The output layer is a dynamically growing repository of neurons with a
limitation of the maximum number of neurons in the repository. The number of input
neurons is fixed and is determined by the user-defined parameter NIsize
, while the maxi-
mum number of output neurons is specified by NOsize
. The continuously incoming input
data of a univariate input data stream is buffered in a sliding window W of size Wsize for
internal processing. The values inside the window are encoded to spike times for the
input neurons using a fixed number of evenly distributed overlapping Gaussian Recep-
tive Fields (GRF) which are further described in Sect. 4.3. For the output layer a simpli-
fied neural LIF model (Schliebs and Kasabov 2013; Wysoski et al. 2008) is applied. In
this model, the postsynaptic potential (PSP) is accumulated at the output neuron based
on the input signals from the preceding neural layer until it reaches its post-synaptic
potential threshold 𝜗 as in Eq. (4). Reaching the PSP threshold causes the output neuron
to fire and its PSP value is reset to 0 (Macia̧g et al. 2021).
Based on the error of the output layer, either a new neuron is added to the existing
output layer or the parameters of an existing neuron are updated (Kasabov 2006). At
first a new output neuron is created for each value xt of the univariate input data stream.
The corresponding weights for the newly created neuron are initialized according to the
spike order of the input neurons as described in Wysoski et al. (2006). Each newly cre-
ated output neuron is then either added as a new instance to the output neuron reposi-
tory or merged with an existing output neuron in the output repository. This behavior is
controlled by a user-defined parameter sim, which is located in the (0, 1] range. For this
purpose, the similarity between the newly created output neuron k and each of the other
output neurons in the repository is calculated. The similarity is defined as the reciprocal
of the Euclidean distance between the weights of the newly added output neuron and the
other output neurons. If the similarity to one of the existing neurons exceeds the prede-
fined threshold value sim, then the newly added neuron will be merged with the most
similar neuron as in Maciąg et al. (2021). If no existing output neuron is similar enough
for the defined threshold sim, the new output neuron is added to the output repository.
If the repository has reached its maximum size, the oldest neuron in the output layer is
replaced by the newly created neuron (Macia̧g et al. 2021).
For anomaly detection a vector of error values is calculated between predicted and
observed values of the window W. Therefore for every predicted scalar value yt and
observed scalar value xt of the window W the absolute difference et = |xt − yt| is calcu-
lated, which results in an error vector 𝐞 of the dimension Wsize
. Based on this vector, the
mean value ̄
xe and the standard deviation s2
e
of the error values of 𝐞 are used to classify xt
either as normal or anomalous. If the difference between et and ̄
xe is greater than 𝜖 ⋅ s2
e
,
where 𝜖 is a user-defined threshold, then xt is classified as an anomaly (Macia̧g et al.
2021).
In summary, each scalar time series value xt goes through the following steps to gen-
erate a prediction and gets classified as normal or anomalous:
1. The input window W is updated with the value xt and the GRF of the input neurons are
initialized.
2. The value xt is encoded by the GRF into spike times (see Sect. 4.3).
3. The coded spike times are then used to determine the spike order of the input neuron.
4. Based on the spike order of the input neuron, the PSP of the output neuron is accumu-
lated in the same order.
5. The output neuron that reaches its PSP threshold first generates the prediction yt.

1384
1 3
6. The new prediction yt is compared to the actual value xt and classified as normal or
anomaly.
7. If no anomaly is detected, the output neuron that generated the prediction yt is corrected
according to the deviation from the input value xt.
8. In parallel to steps 6 and 7, a new output neuron is generated independently of the
anomaly classification. This neuron is then merged or added to the repository depending
on the similarity threshold sim as described before.
Originally, the model by Maciąg et al. (2021) with the name OeSNN-UAD according
to Online evolving Spiking Neural Networks for Unsupervised Anomaly Detection was
designed for univariate data processing only. For anomaly detection in multivariate time
series, one instance of the model can be executed per dimension, but then no correlation
between the dimensions is considered as shown in Sect. 5. We therefore develop an appro-
priate measure to improve the processing of multivariate data as described in Sect. 4.3. Our
focus is primarily on the runtime and real-time capability as well as the general perfor-
mance of the detection of different types of anomalies. When talking about detection it is
important to look at the the correctly classified outliers, the unrecognized outliers and the
data instances which are incorrectly classified as outliers. We deal with this in Sect. 5.4,
where we define the F1 score as an appropriate metric to compare our results to other result
in the literature. Additionally, we improve the anomaly detection by extending the model
with an anomaly probability score (see Sect. 4.5) and reduce the computing costs through
an improved learning approach which is described in Sect. 4.4.
4.3 Efficient encoding of multivariate data with Gaussian Receptive Fields
The biological inspired structure of receptive fields and especially the Gaussian Receptive
Fields plays an elementary role in the efficient processing of data in SNN. In the biologi-
cal context, a receptive field is defined as a spatially limited area of interconnected sensory
receptors that convert incoming visual, auditory or similar stimuli into electrical stimulus
potentials (Lindeberg 2013). The incoming visual stimuli are encoded by receptive fields
into electrical signals that are further processed by subsequent neuron layers. Due to the
sparse interaction of receptive fields with subsequent neurons, this coding can be used
to efficiently process the information in the subsequent neuron layers (Lindeberg 2013).
This biological concept is applied in different types of artificial neural networks, such as
the Convolutional Neural Networks (Goodfellow et al. 2016) or the Spiking Neural Net-
works (Macia̧g et al. 2021; Panuku and Sekhar 2008; Hopkins et al. 2018).
The encoding of the input data is performed in Maciąg et al. (2021) using a one dimen-
sional Gaussian Receptive Field, where multiple Gaussian distributions are placed equally
over the input window W. This approach however limits the algorithm to one dimensional
input data. The parallel execution of several OeSNN instances for processing multivari-
ate data in each dimension entails several limitations which affect the performance of the
model in terms of runtime and the identification of complex outliers. In order to eliminate
these limitations or to minimize them in their manifestation, we examine in the following
an efficient modeling technique for multidimensional Gaussian Receptive Fields. It is sum-
marized at the end of this subsection with points 1. to 4.
We use multidimensional Gaussian distribution functions on multidimensional clus-
ters Ci
, which we obtain from the input data via a k-Means clustering. For the two-
dimensional case the clustering algorithm is illustrated in the upper left-hand side of

1385
1 3
Fig. 4. We do the clustering based on the work of Panuku and Sekhar (2008) in which
they model a defined number of multivariate Gaussian distributions over the incom-
ing data. The placement of the receptive fields differs here from the one-dimensional
variant of the OeSNN-UAD architecture. In contrast to the OeSNN-UAD architecture,
the placement of the receptive fields is not evenly distributed over the incoming data,
instead it is placed specifically in those regions where clusters appear (cf. Fig. 1). This
approach can significantly reduce the required number of receptive fields on the input
data, which in turn yields a positive effect on the runtime of the algorithm.
The excitation factor of the following input neurons is calculated for each input neu-
ron in the same manner as in the one-dimensional approach via the normal distribu-
tion function of the respective cluster Ci
, as it is illustrated in Fig. 4 in the upper right-
hand side. It is defined, as shown in Eq. (5), by the center 𝛍i ∈ ℝn
and the covariance
Σi ∈ ℝn×n
of the cluster Ci.
From the excitation factor fCi
(x) the spike time ti is calculated by
as it is shown in Fig. 4 by moving from the upper right-hand side to the lower left-hand
side. However, the variant as proposed by Panuku and Sekhar (2008) is primarily intended
for an offline setting, in which the receptive fields are initialized in an initial training phase.
In order to preserve the online character of the anomaly detection component, both the
k-Means clustering and the calculation of the covariance of the clusters for each incoming
data point are calculated recursively.
For the clustering of the incoming data, an online-capable version of the Lloyd
k-Means algorithm (King 2012) is used, which can perform the adjustment of the clus-
ters in a streaming based application with constant runtime. Here, all Ci clusters are
initiated by the incoming data and all further data points are assigned to the existing
cluster with the smallest Euclidean distance.The associated cluster center 𝛍i as well as
the covariance Σi is recursively estimated for each data point 𝐱t belonging to the cluster
using Eqs. (7) and (8).
(5)
fCi
(x) = exp
[
−
1
2
(
𝐱 − 𝛍i
)T
Σ−1
i
(
𝐱 − 𝛍i
)]
(6)
ti = 1 − fCi
(x)
Fig. 1 Visualization of the estimated clusters (k = 2) using the online version of the GRF algorithm for a
two-dimensional time series of the Numenta dataset with anomaly

1386
1 3
In summary the calculation of the encoding of multivariate data streams is done via the
following iterative algorithm. It differs from Maciąg et al. (2021) not only in the usage
of multivariate Gaussian distributions but even more by the clustering of the data points,
which makes it possible to reduce the number of Gaussian distributions and obtain better
results:
1. Calculate a predefined number of multivariate clusters Ci with center 𝜇i and covariance
matrix Σi via k-Means clustering with the data already obtained.
2. Place a multidimensional Gaussian distribution with mean 𝜇i and covariance matrix Σi
over each of the clusters Ci.
3. For each incoming data point xt calculate the excitation factors fCi
(xt) for each of the
cluster Ci according to Eq. (5).
4. Assign each incoming data point xt to the nearest cluster Ci and update the mean 𝜇i and
covariance matrix Σi of this cluster as in Eqs. (7) and (8) and start again with 1.
In order to verify the effectiveness of the changes made for the online-capable opera-
tion in comparison to the standard algorithms, we execute both versions of the algo-
rithms for a signal with 10,000 data points and compare the results.
In a subsequent step, as shown in Fig. 2, we determine the absolute distances (Man-
hattan distance) of the vector or matrix components between the cluster center 𝛍i and
the estimated covariance matrices Σi for the time series shown in Fig. 1.
We observe that both the cluster center and the covariance matrices asymptotically
converge to the results of the offline algorithm. The estimated cluster center converge
significantly faster than the covariance matrices. Nevertheless, in the evaluation of the
algorithm, we could not detect any negative influence due to the slightly inaccurate
covariance estimation.
(7)
𝛍i =𝛍i,l = 𝛍i,l−1 +
1
l
(
𝐱t − 𝛍i,l−1
)
(8)
Σi =Σi,l =
(
1 −
1
l
)(
Σi,l−1 +
1
l
(
)(
)T
)
Fig. 2 Comparison of the convergence properties of cluster centers (a) and cluster covariances (b) of the
online-capable adaptation and the offline reference algorithm according to Panuku and Sekhar (2008)

1387
1 3
4.4 Learning capabilities with dynamic weight adaption
The architecture by Maciąg et al. (2021) uses a learning method adopted from Wysoski
et al. (2006) in order to calculate the networks weights. This online learning method
uses the ranking of the occurring spikes to update the weights. The weight change is
calculated with
where 𝛥wji is the change in weight wji between neuron j in the input layer and neuron i in
the output layer. The constant mod represents the modulation factor, which is in the range
of (0, 1). The order(j) parameter corresponds to the index for the neuron j in a list sorted
by spike time. The exact time information is discarded and not included in the calculation
of the weights. Wang et al. (2017) published an improved Rank-Order-Based learning pro-
cedure for SNN, which is called SpikeTemp. In SpikeTemp the spike time is used directly
to update the weights. Equation (10) describes the change in the weights wji of the synapse
that connects a neuron j to an output neuron i, where tj represents the spike time of the
neuron j and 𝜏 is a constant scaling factor used as a hyperparameter (see Sect. 5.3) (Wang
et al. 2017).
In this learning procedure, the available time information is used directly and not only
the order of the spike times as in Eq. (9), so that the time interval influences the weight
change. When using the Rank-Order-Based approach of Wysoski et al. (2006) the learn-
ing algorithm only provides a constant weight update. Therefore, the postsynaptic potential
(PSP) for this output neuron is always identical for an input pattern that may have differ-
ent spike times but produces the same fire order. With the SpikeTemp approach, different
spike times always lead to different weight changes and a different PSP as determine in Eq.
(11). Consequently, the weight changes and the PSP correlate better with the input pattern,
which contributes to improved learning performance. Since the SpikeTemp approach also
eliminates the need to order all spikes in a window, it also reduces the computational effort
(Wang et al. 2017).
The following steps describe the customized learning procedure with SpikeTemp for our
online evolving SNN model. It is also illustrated in the lower right hand side of the over-
view picture Fig. 4 explaining our model.
1. For each input value an output neuron is generated as described in Maciąg et al. (2021).
However, the weights to the input neurons are now initialized to a constant value. Wang
et al. (2017) determined the experimental value of 0.1 for a SNN classification model.
We change this constant initialization factor for the online evolving SNN to a hyperpa-
rameter in the model, using 0.1 as a reference value (see Sect. 5.3). The weight change
from Eq. (10) is then added to the initial base weight.
(9)
𝛥wji = modorder(j)
(10)
𝛥wji = exp
(
−
tj
𝜏
)
(11)
PSP(i, t) =
∑
j∈[1..N]
wji ⋅ exp
(
−
tj
𝜏
)

1388
1 3
2. The PSP for an additional neuron in the output layer is now calculated using Eq. (11).
The calculation of the neuron fire threshold remains unchanged and is adapted from the
architecture of Maciąg et al. (2021).
Fig. 3 Calculation of the outlier score: Input and the prediction by the OeSNN model (left); The resulting
quadratic reconstruction error (middle); The Outlier score given by the complement of the Q-function on
the squared reconstruction error (right)
dynamic output
neuron repository
Gaussian receptive field
cluster
excitement
-
where
cluster
cluster
cluster
-
coordinate
- coordinate
update clusters for
each incoming
Online evolving Spiking Neural Network
input
neurons
update
repository
No anomaly
detected
classify
anomaly
error
correction
spike
times
-
cluster
Fig. 4 Illustration of multivariate online evolving SNN model (OeSNN-D)

1389
1 3
3. The addition of the new output neuron to the output neuron repository remains
unchanged to Maciąg et al. (2021). If the similarity to one of the existing output neurons
is greater than a predefined threshold, the newly added output neuron is merged with
the most similar output neuron, otherwise the output neuron is added to the repository.
4.5 Anomaly detection using a continuous outlier score
The architecture of Maciąg et al. (2021) uses the difference between the prediction and
the actual value and compares the deviation to the previously determined deviations. As
soon as this exceeds a user defined threshold value, the data point is classified as an outlier.
However is not unusual to have occasional jumps in a time series, which lead to prediction
errors, because of a slight time shift between prediction and input data (see Fig. 3). To
handle these scenarios we replace the outlier detection function by the Numenta scoring
function of Ahmad et al. (2017). This makes it possible to determine for each data point by
use of the reconstruction error a probability with which a data point belongs to the outlier
category. For this purpose, the squared reconstruction error (see Fig. 3 middle) for each
step t is assumed to be a representation of a continuous normal distribution function. To
calculate the distribution the continuous mean 𝜇t and variance 𝜎t are calculated based on
the squared reconstruction error. The exact procedure is shown in Fig. 3. The green area
marks the initialization phase of the algorithm.
The left picture shows the time series and the prediction of our new algorithm. Here the
anomaly consists of a missing amplitude starting at the index 1200. From this the squared
reconstruction error is calculated, which is shown in the center. The outlier probability is
given by the complement of the Q-function on the squared reconstruction error as shown
in Eq. (12):
The variable ̃
𝜇t represents the mean value of the squared reconstruction error of a defined
time window W. The mean 𝜇t and the variance 𝜎t are calculated using the Welford online
algorithm (Welford 1962). The outlier score in Eq. (12) is calculated for each dimen-
sion separately on the corresponding reconstruction error and is then combined into one
score by just taking the highest value of the outlier score over all dimensions at time step
t. Replacing the outlier detection with the Numenta scoring function allows our newly
adapted model to output a continuous outlier probability, whereas the architecture of
Maciąg et al. (2021) only outputs a binary classification of outliers. However, when com-
paring our algorithm with other anomaly detection algorithms in Sect. 5 we clearly need
to identify whether a single multivariate data point belongs to an anomaly or not, so that a
binary classification is needed. We therefore classify every data point with an outlier score
above the threshold of 0.8 as an outlier and every data point with an outlier score below 0.8
not as an outlier as it is also indicated in Fig. 3 (right).
Papadimitriou et al. (2005) use the number of eigencomponents k as an anomaly indica-
tor in the SPIRIT algorithm to determine the outlier class. To achieve a better comparabil-
ity between the reference algorithm and the modified OeSNN, the outlier score in Eq. (12)
is also used in the SPIRIT data processing pipeline. For this purpose, the reconstruction
error per dimension is used as an input. The outlier scores of the individual dimensions are
then merged into one outlier score using the respective maximum value, which makes both
models easily comparable.
(12)
Lt = 1 − Q
(
̃
𝜇t − 𝜇t
𝜎t
)
.

1390
1 3
5 Experimental evaluation
5.1 Overview of models
For the following evaluation of the performance of algorithms, we combined the adjust-
ments in Sects. 4.3, 4.4 and 4.5 of the initial algorithm according to Maciąg et al. (2021)
incrementally as shown in Table 1. Therefore we address all models based on the initial
OeSNN-UAD algorithm with prefix OeSNN. In addition, we add letters A-D consecutively
according to the integrated extensions. Here the model OeSNN-A represents the version
in the paper of Maciąg et al. (2021), which is extended with each following letter by the
adaptations shown in Table 1. For a general overview of our final model OeSNN-D we also
refer to Fig. 4.
The models were implemented with Cython (Behnel et al. 2011). Wrapping the exter-
nal C++ library as a Python extension allows an accelerated execution compared to a
Python native implementation, which is especially beneficial for online and real-time data
processing.
5.2 Evaluation benchmark
We use the Numenta Anomaly Benchmark (NAB) (Ahmad et al. 2017) to analyze the pre-
sented modifications of the OeSNN and to identify strengths and weaknesses of the adjust-
ments. NAB consists of over 50 labeled real and artificial time series. All data including
a comprehensive documentation are available as open source.1
Much of the data consists
of real time series data and comes from a variety of sources such as AWS server metrics,
Twitter volumes, web click metrics, traffic data and more. These time series are mainly uni-
variate records, which means that they are not suitable for our multivariate comparison and
evaluation. Furthermore, the synthetic dataset contains only a few anomaly types. In order
to allow a differentiated comparison of the performance of the different OeSNN modifica-
tions on the largest possible number of different outlier species, we added additional outlier
types to the synthetically generated part of the Numenta dataset. Our extended benchmark
data set, which is also shown in appendix A.1, contains outliers of the following catego-
ries: signal dropouts, signal drift, lower amplitude, higher amplitude, cycle dropout, single
peaks, change of frequency, increase in noise and increase in amplitude. For the evaluation
of multivariate time series the synthetic datasets are combined so that one outlier is present
in one dimension, while the other dimensions use the anomaly-free synthetic signal.
Table 1 Model overview
Function OeSNN-A OeSNN-B OeSNN-C OeSNN-D
Numenta outlier score (see Sect. 4.5) No Yes Yes Yes
Multivariate GRF (see Sect. 4.3) No No Yes Yes
SpikeTemp learning (see Sect. 4.4) No No No Yes
1
https://github.com/numenta/NAB.

1391
1 3
5.3 Hyperparameter optimization
In the literature a wide range of different algorithms for the optimization of hyperparam-
eters is presented (Yu and Zhu 2020; Bergstra et al. 2011). Many of these algorithms,
such as grid search, use a relatively naive approach for the selection of hyperparameter
optimization, which means that the effort for the optimization of the given parameters for
high dimensional optimization problems can only be realized with massive computational
resources. Furthermore, since the number of outliers in both the given artificially generated
and real data belong to the minority class, it is not possible to fall back on a large data pool
for hyperparameter optimization here. Therefore, it cannot be guaranteed that there is no
overfitting of the parameters to a certain type of outlier, which could influence the generali-
zation ability of the algorithms. In addition, in many real-world applications, outliers of the
signal are not present for optimization, which presents further difficulties.
Due to the high number of hyperparameters of the OeSNN (see Appendix A.2) as well
as the low number of test and validation data, we present in the following sections a tech-
nique to effectively select the hyperparameter candidates as well as an approach to opti-
mize the hyperparameters without prior knowledge of the signal-specific outlier types.
5.3.1 Generation of validation data
In order to avoid overfitting of the model due to the small size of the data instances of the
NAB dataset and to enable a method for optimizing the algorithm without prior knowledge
of the signal-specific outlier types we investigate a method of optimization using only an
artificially generated anomaly. For this purpose, we replace in one dimension of the anom-
aly-free signal a window slightly larger than one cycle with an artificially generated noise
as illustrated in Fig. 5.
For the optimization of the model, we use the reconstruction error of the predicted sig-
nal on the artificially generated anomaly signal and the anomaly-free signal as a quality
measure. In comparison to an optimization with only one real outlier type, this method
provides consistently comparable or better generalization results (see A.5).
5.3.2 Efficient hyperparameter selection
In order to reduce the required number of hyperparameters to be tested and to minimize the
required time for hyperparameter optimization we used the tree-structured parzen estima-
tor (TPE) approach (Bergstra et al. 2011, 2013). We implement the optimization with the
Python framework Optuna (Akiba et al. 2019). Here we optimize the hyperparameters in
1500 trials as shown in Table 2, of which 20% are initialized with random parameters. For
the selection of the next parameters, 50 possible parameter configurations are randomly
chosen per iteration, which is represented by the EI-candidates value in Table 2. On the
Fig. 5 Signal of the Numenta Anomaly Benchmark without anomaly (left) and the corresponding generated
data (right) with inserted artificial noise, which is used for hyperparameter optimization

1392
1 3
basis of these configurations the expected improvement of the model is calculated. We run
all iterations of the optimization algorithm sequentially on one thread to obtain reproduc-
ible results. We also initialize a random generator, which is used for the selection of the
parameter candidates, with a constant seed. We consider the selection of the optimal hyper-
parameters (
𝜃∗

) as an optimization problem with constraint (cf. Eqs. (13) and (14)). The
aim is to minimize the mean squared error between the reconstruction output (
̂
𝐲i
) of the
OeSNN algorithm with the artificially generated outlier ( ̃
𝐱i
) as input and the original signal
without outlier (
𝐱i
) under the constraint that for less than1% of the data instances k none of
the output neurons fires. Constraint Eq. (14) is necessary, since otherwise the best hyperpa-
rameters are found by simply reproducing the original data, which leads to an overfitting.
Instead of1% any small value can be used to prevent overfitting.
We perform this optimization task for each of the described OeSNN for the two-dimen-
sional case. The respective search space for the hyperparameters can be looked up in the
appendix A.2. The optimization results are then transferred to all other dimensions. Since
the defined optimization criterion does not allow direct optimization of the window size
parameter of the outlier score, we adjusted this parameter manually after the optimization.
Here we selected a value as a trade off between a high sensitivity to changes in the signal
and an overall low variance of the score.
5.4 Model evaluation
To evaluate the algorithms for detecting anomalies in a streaming context we perform four
quantitative benchmarks for the created variations of the OeSNN algorithm. They consist
of a runtime analysis of the models, which will be examined in the following with regard
to an increasing number of data points as well as an increasing number of dimensions. In
addition, we investigated the anomaly detection with regards to the suitability for detecting
anomalies in high-dimensional data as well as the detection of various types of anomalies.
The exact configuration of the benchmarks is shown in Table 3.
To estimate the performance of the OeSNN in the context of online anomaly detection
algorithms we used the SPIRIT algorithm according to Papadimitriou et al. (2005) as refer-
ence algorithm. In order to compare the OeSNN-D algorithm with deep learning methods
as in Munir et al. (2019), we also evaluated the OeSNN-D algorithm on a different dataset,
(13)
𝜃∗= arg min𝜃
1
k
∑k
i=1
�
̂
𝐲i − 𝐱i
�2
= arg min𝜃
1
k
∑k
i=1
�
OeSNN
�
𝜃, ̃
𝐱i
�
− 𝐱i
�2
(14)
subject to tnot fired 0.01 ⋅ k
Table 2 Parameters of the TPE algorithm for hyperparameter search
Parameter Description Value
Trials Number of iterations of the TPE algorithm 1500
Startup-trials Number of initially randomly determined parameters 300
EI-candidates Number of candidates for the calculation of the expected improve-
ment
50

1393
1 3
the Yahoo Webscope dataset and compare it to the results obtained in Munir et al. (2019).
For all experiments the hyperparameters, as determined in the previous section, are pro-
vided in the appendix A.3.
Selecting a suitable metric for the evaluation of the anomaly recognition rate of the
algorithms is an important component. The goal is to select a suitable metric that is invari-
ant with respect to an imbalance between normal data points and outliers and provides a
suitable basis for comparing the model with other algorithms. For the quantitative determi-
nation of the performance of the algorithms we use the F1 score [Eq. (15)] on the bench-
mark datasets. Although it is used much less frequently in the scientific literature than the
accuracy score (Hossin and Sulaiman 2015), it is especially suitable for classifications with
unequally weighted classes (Jeni et al. 2013). In addition, the F1 score can be found in a
number of other publications in the field of anomaly detection (Macia̧g et al. 2021; Däu-
bener et al. 2019; Munir et al. 2019), which facilitates a comparison of the algorithms:
The F1 score is calculated on the basis of the fraction of correctly predicted positive
instances from all positive predictions P (Precision) and the fraction of correctly predicted
positive instances from all positive instances R (Recall) (Hossin and Sulaiman 2015):
where tp are the correctly classified outliers, fn unrecognized outliers and fp data instances
which were incorrectly classified as outliers.
5.5 Experimental determination of runtimes
To analyze runtime, we make use of an experimental evaluation instead of a theoreti-
cal analysis of the runtime of the algorithm via O-notation. To ensure the comparability
of the results, we executed the tests sequentially one after the other on a defined system
(i9-9900k, 32 GB DDR4 RAM, Ubuntu 20.04.1 LTS) without any additional load that
could have a negative impact on the results. In order to obtain the most precise results
possible, the selection of the timer module also has an important role. Special attention
should be paid to a high temporal resolution. For this purpose, we use the process_time_ns
(15)
F1 = 2 ⋅
P ⋅ R
P + R
(16)
Precision(P) =
tp
tp + fp
(17)
Recall(R) =
tp
tp + fn
Table 3 Configuration of benchmarks to evaluate the algorithms
Experiment Dataset Data points Dimension
Runtime data points Randomly generated 103
–106 2
Runtime dimensionality NAB ∼ 2500 2–4
Anomaly detection dimensionality NAB ∼ 2500 2–4
Anomaly detection types NAB ∼ 2500 2–4

1394
1 3
function from standard Python library (Stinner 2017) which has the highest temporal reso-
lution with 1 nano second. To compensate the inaccuracies due to the parallel occurring
load or further inaccuracies in the determination of the runtime, all experiments for the
analysis of the running time are performed with a sample size of m = 20.
5.5.1 Influence of number of data points
Based on the tests performed regarding the change in runtime with increasing number of
data points, it can be determined that the OeSNN-D version of the algorithm scales best
with new incoming data points when only 2 clusters are required for the distribution of the
multivariate Gaussian recipe fields as shown in Fig. 6. The runtime of the algorithm for
106
data points is only 40% of that of OeSNN-A. Also compared to the SPIRIT reference
algorithm as well as the other variants of the OeSNN, the runtime of the algorithm is in a
very good range. Especially noticeable is the improvement of the OeSNN from variant C to
variant D. The Spiking Neural Network with preceding multivariate GRF requires a mul-
tiple of the runtime of the final model. This reduction of the runtime is described by Wang
et al. (2017) and is also reinforced by the reduced number of input neurons.
Moreover, we observe that the adapted scoring system, which enables a real-value clas-
sification of outliers, can achieve a significant improvement in runtime (cf. OeSNN-A and
OeSNN-B).
5.5.2 Influence of dimensionality of the input data
As the dimensionality of the data increases, the number of processing points also increases,
which makes preserving real-time capability almost impossible. Among the algorithms as
well as their variations we could prove that the variants (OeSNN-A; OeSNN-B) that allow
processing of multivariate data only by executing several instances with increasing dimen-
sionality do not benefit from the correlation of the data dimensions (see Fig. 7). With the
OeSNN-B variant of the algorithm, which uses an adapted online-capable variant of the
Numenta outlier score. However, we could observe a significant improvement compared to
variant OeSNN-A.
All other variants of the algorithm (OeSNN-C, OeSNN-D), which are based on cod-
ing multivariate data using the multidimensional GRF, scale significantly better with
Fig. 6 Runtime of the investigated algorithms with increasing number of data points

1395
1 3
increasing dimensionality as can be seen from Fig. 7. The runtime of the OeSNN-D from
the one-dimensional to the twenty-dimensional input signal increases by a factor of about
3.5, whereas the runtime of the initial OeSNN-A increases by a factor of 21.
5.6 Anomaly detection of different algorithms
In addition to the runtime, a high detection rate of anomalies in time series is a fundamen-
tal requirement. In the following, this will be analyzed and discussed with regard to the
general detection rate for several dimensions and the average detection rate of different
types of anomaly in the NAB dataset. For the evaluation, we used the F1 score described
in Sect. 5.4. The complete results of the evaluation are provided in Appendix A.5. Further-
more in order to evaluate the new OeSNN-D algorithm on a different dataset we inves-
tigate in Sect. 5.6.3 its performance on the Yahoo webscope dataset, which consists of
four groups of benchmark datasets A1, A2, A3, A4. These datasets are also investigated in
Munir et al. (2019) with a deep learning algorithm.
5.6.1 Anomaly detection and increasing dimensionality of the data
Our analysis show that the average F1 score of the OeSNN-D model for the tested
NAB dataset with dimensions (2–4) gives better results than all compared models (see
Table 4). Remarkable is the increase of the average F1 score compared to the initial model
Fig. 7 Runtime of the investigated algorithms with increasing number of dimensions
Table 4 Average F1 score of the
models for detecting anomalies
across all anomaly types of the
NAB dataset
Highest F1 score values are given in bold
Dimension F1 score over all time series of the NAB
SPIRIT OeSNN-A OeSNN-B OeSNN-C OeSNN-D
1 – 0.22 0.47 0.26 0.47
2 0.42 0.22 0.44 0.61 0.61
3 0.32 0.22 0.44 0.60 0.62
4 0.31 0.22 0.44 0.57 0.62
Average 0.35 0.22 0.45 0.51 0.58

1396
1 3
(OeSNN-A) by 0.36, due to the different outlier score calculation. Furthermore, we show
that by coding the multidimensional data for the models OeSNN-C and OeSNN-D, the rec-
ognition rate of the algorithms can be further improved.
Comparing the models OeSNN-B, OeSNN-C and OeSNN-D as in Table 4, we notice
that the F1 score increases with each version. The results in Table 4 are in some cases
also astonishing since with the implementation of the GRF the higher dimensional cases
do have a better F1 score than the one dimensional case. The reason is that several out-
lier types as for example cycle dropouts and change of frequency perform really bad when
calculating the F1 score in the one dimensional case for OeSNN-C and OeSNN-D (see
Sect. A.4). But they perform really good in the higher dimensional case, when the anomaly
occurs only in one and not in all dimensions. In this case the multidimensional data points
of the anomalies lie in another region than the normal data points, which makes it easier
for the GRF to detect them. This of course shows that our OeSNN-D model has its real
strength in the multivariate case.
5.6.2 Detection of different outlier types in the NAB dataset
As shown in Table 5, the OeSNN-D model achieves the highest F1 score for most anomaly
types in the NAB dataset. This includes all categories except for the outlier types cycle
dropouts, increase in noise and increase in amplitude (see Appendix A.5). However, for
the categories cycle dropouts and increase in amplitude, the model is with a deviation of
0.01 and 0.05, respectively, very close to the highest achieved F1 score. We identify a defi-
cit of the OeSNN-D model compared to the reference algorithm (SPIRIT) in the outlier
category increase in noise.
As previously with the analysis of the recognition rate in Sect. 5.6.1, we also see a step-
by-step improvement with increasing dimensionality of the data for the models OeSNN-C
and OeSNN-B, whereby we could also prove the positive influence of the individual modi-
fications on the performance of the final model (OeSNN-D). For the case of increase in
noise the principal component in the SPIRIT algorithm is not able to converge, therefore
the algorithm detects an anomaly. Whereas in the OeSNN algorithms the increasing noise
oscillates around the mean value of a Gaussian receptive field which leads to similar spike
Table 5 Average F1 score of the models for detection of the anomaly types of the NAB dataset averaged
over the dimensions 2-4
Anomaly F1 score over all dimensions of a time series
SPIRIT OeSNN-A OeSNN-B OeSNN-C OeSNN-D
Signal dropouts 0.65 0.01 0.62 0.54 0.78
Signal drift 0.41 0.02 0.51 0.66 0.73
Lower amplitude 0.15 0.60 0.62 0.69 0.95
Higher amplitude 0.35 0.87 0.78 0.90 0.90
Cycle dropout 0.12 0.26 0.57 0.67 0.66
Single peaks 0.00 0.09 0.00 0.16 0.16
Change of frequency 0.50 0.16 0.46 0.46 0.51
Increase in noise 0.92 0.00 0.00 0.09 0.09
Increase in amplitude 0.04 0.01 0.49 0.42 0.44

1397
1 3
values such that no anomaly is detected.This also illustrates the different nature of both
algorithms. Whereas SPIRIT detects outliers, whenever the principal component of a mul-
tivariate data stream changes, our OeSNN-D algorithm detects outliers if the clusters of
multidimensional datasets change in length or form.
5.6.3 Anomaly detection in the Yahoo Webscope dataset
The main focus of the newly developed anomaly detection algorithm OeSNN-D is to detect
online anomalies in multivariate time series of periodic or seasonal processes i.e. in pro-
duction processes. The results in Sect. 5.6.2 show that the performance of the OeSNN-D
algorithm is very good if structural anomalies are considered. In this section we evalu-
ate the Yahoo Webscope dataset, a publicly available data set released by Yahoo Labs. It
consists of 367 real and synthetic time series with point anomaly labels. Each time series
contains 1420 to 1680 instances. It is divided into four sub-benchmarks namely A1 Bench-
mark, A2 Benchmark, A3 Benchmark and A4 Benchmark. Most of the anomalies in the
Yahoo Webscope datasets are point anomalies, which means there is only one anomalous
time series index in a certain neighbourhood. The Yahoo Webscope dataset was also con-
sidered in Munir et al. (2019), where deep learning methods were applied to detect anoma-
lies. We evaluated the anomalies of the Yahoo Webscope dataset as it was done in Munir
et al. (2019) by determining the hyperparameter for every sub benchmark A1, A2, A3, A4
seprately (see Appendix A.3) the results can be seen in Table 6.
Although the F1 score of the OeSSN-D algorithm can not compete with the deep learn-
ing models of DeeptAnT, it is obtains most of the time better results than other none deep
learning algorithms. Since the main focus of our algorithm lies in the online detection of
anomalies without a training phase, the results are very promising also for this data set.
However, since most of the anomalies in the Yahoo Webscope dataset are point anom-
alies, a separate algorithm for point anomalies can be developed and, in the best case,
put into parallel operation with the OeSNN-D algorithm. Most point anomalies can be
detected by an algorithm which calculates the differences of successive times series values
𝛥xt = xt − xt−1 in a certain sliding window w1 with length |w1| and compare them to the
directly following differences of a very small sliding window w2 of length|w2| . Thus,
Table 6 Average F1 score for the Yahoo Webscope dataset in comparison with the results in Munir et al.
(2019)
Table entries marked with * are the results of Munir et al. (2019)
Data
category
Yahoo
EGADS*
Twitter
Anomaly
Detection,
𝛼 = 0.05*
Twitter
Anomaly
Detection,
𝛼 = 0.2*
DeepAnT
(CNN)*
DeepAnT
(LSTM)*
OeSNN-D
A1 0.47 0.48 0.47 0.46 0.44 0.405
A2 0.58 0 0 0.94 0.97 0.451
A3 0.48 0.26 0.3 0.87 0.72 0.110
A4 0.29 0.31 0.34 0.68 0.59 0.147

1398
1 3
A point anomaly is detected whenever
where 𝜎w1
denotes the standard deviation and w1 the mean of w1
. The parameters 𝜖, |w1|and
|w2| are hyperparameters, which are set to 𝜖 = 4.5 , |w1| = 32 and |w2| = 4
. Evaluating the
Yahoo Webscope with this algorithm detects most of the point anomalies. The reuslts are
shown in Table 7.
6 Conclusion
In this paper we introduce an online capable algorithm to efficiently detect anomalies in
multivariate time series. We therefore use an online Spiking Neural Network architecture
with a data efficient technique for coding higher-dimensional data using Multidimensional
Gaussian Receptive Fields (GRF) as well as advanced learning methods for improved mul-
tidimensional anomaly detection.
In the evaluation of several versions of multidimensional online Spiking Neural Net-
works (OeSNN-A; OeSNN-B; OeSNN-C; OeSNN-D) using the Numenta Anomaly
Benchmark (NAB) and the Yahoo Webscope dataset we prove that several modifications
of the initial OeSNN algorithm according to Maciąg et al. (2021) significantly improve the
general suitability for outlier detection in higher dimensional (
n ≥ 2 ) data and the detection
of specific outlier types while requiring consistently less runtime than the initial algorithm.
In addition, we determine a better detection result for almost all outlier types compared to
the reference SPIRIT algorithm. Our new algorithm shows deficits only in the outlier types
increase in noise and individual peaks. Individual peaks or point anomalies are hard to
detect with our algorithm, but they can be detected with a different algorithm as we showed
in Sect. 5.6.3.
To assess the suitability of the adapted OeSNN further, we will investigate more com-
plex signal types from different application domains. Thus, we also check whether our
w1 ={𝛥xt−|w1|−|w2|, 𝛥xt−|w1|−|w2|+1, ..., xt−|w2|} and
w2 ={𝛥xt−|w2|, 𝛥xt−|w2|+1, ..., xt}.
|𝛥xt − w1| ≥ 𝜎w1
⋅ 𝜖 and |
∑
v∈w2
v| |𝜎w1
⋅ (𝜖 − 0.5)|,
Table 7 Average F1 score for the point anomaly detector on the Yahoo Webscope dataset in comparison
with the results in Munir et al. (2019)
Table entries marked with * are the results of Munir et al. (2019)
Data
category
Yahoo
EGADS*
Twitter
Anomaly
Detection,
𝛼 = 0.05*
Twitter
Anomaly
Detection,
𝛼 = 0.2*
DeepAnT
(CNN)*
DeepAnT
(LSTM)*
OeSNN-
D
and Point
Anomaly
Detector
A1 0.47 0.48 0.47 0.46 0.44 0.405
A2 0.58 0 0 0.94 0.97 0.543
A3 0.48 0.26 0.3 0.87 0.72 0.639
A4 0.29 0.31 0.34 0.68 0.59 0.609

1399
1 3
algorithm meets the requirements of Wu and Keogh (2021). The authors there question
the usage of deep learning models for anomaly detection in time series. That is why we
believe that spiking neural networks and specifically our OeSNN-D model might be one
of the best ways to detect anomalies, which are not trivial and can be weed out with two
lines of code, in multivariate streaming times series. As pointed out in this paper and in
Wu and Keogh (2021) as well, a lot of the point anomalies of the Yahoo Webscope data-
set can be detected with a really simple algorithm. The focus in this paper is however to
detect structural changes in multivariate times series and also to categorize the different
type of anomalies whether they are detectable by our algorithm or not. In order to further
strengthen the online capability of our algorithm the number of required hyperparameters
must be reduced in further work or some new insights for transferring existing hyperpa-
rameters between different application domains must be established.
A Appendix
A.1 Signal categories of the extended Numenta Anomaly Benchmark
For adding additional anomalies which are not contained in the original NAB dataset, we
used the time series of the NAB dataset and modified it in an adequate way. The time series
signal dropouts, higher amplitude, lower amplitude and cycle dropout are from the original
NAB dataset. For signal drift we added a linear function to the original time series from
a certain time step onwards, for single peaks we added a few peaks to the time series, for
change of frequency we divided the frequency of the original NAB time series by two from
a certain time step onwards, for increase in noise we added an additional gaussian noise to
the original NAB dataset and for increase in amplitude we added a rectangle function with
increasing amplitude from a certain time point onwards (Fig. 8).
Fig. 8 Signal categories of the extended Numenta Anomaly Benchmark

1400
1 3
A.2 Hyperparameter search space
See Table 8.
A.3 Determined hyperparameters for the evaluation of the datasets
See Tables 9, 10, 11, 12, 13 and 14.
Table 8 Hyperparameter search space with p as number of data points for a signal period
Parameter Search Description
Space
Wsize [NIsize, 4p] Input window size
InitPhase [NIsize, 4p] Initialization time steps
NIsize [1, p) Number of input neurons
NOsize [1, p) Maximum number of output neurons
sim (0, 1] Similarity threshold value to merge output neurons
mod (0, 1) Modulation factor for the Rank-Order-Based learning
C [0, 1] Output neuron activation factor
errorCorrection [0, 1] Correction factor for the deviating prediction
anomalyFactor (1, 20) Threshold value to classify an anomaly
scoreWindowSize [1, 4p] Window size for the Numenta outlier score
𝜏 [0.1, 10] Time constant of the SpikeTemp learning approach
weightBias [0, 0.5] Initial bias on weights between neurons
Table 9 Hyperparameters on the modified NAB dataset for the OeSNN-A model
Parameter Value Description
Wsize 157 Input window size
NIsize 13 Number of input neurons
NOsize 74 Maximum number of output neurons
sim 0.42 Similarity threshold value to merge output neurons
C 0.26 Output neuron activation factor
mod 0.63 Modulation factor for the Rank-Order-Based learning
errorCorrection 0.94 Correction factor for the deviating prediction
anomalyFactor 8.93 Threshold value to classify an anomaly
Table 10 Hyperparameters on the modified NAB dataset for the OeSNN-B model
Wsize 165 Input window size
scoreWindowSize 10 Window size for the Numenta outlier score

1401
1 3
Table 11 Hyperparameters on the modified NAB dataset for the OeSNN-C model
InitPhase 59 Initialization time steps
Table 12 Hyperparameter on the modified NAB dataset for the OeSNN-D model
InitPhase 97 Initialization time steps
𝜏 3.29 Time constant of the SpikeTemp learning approach
weightBias 0.25 Initial bias on weights between input and output neurons
Table 13 Hyperparameter on the
Yahoo Webscope dataset for the
OeSNN-D model
Parameter A1 A2 A3 A4
InitPhase 200 400 300 350
NIsize 65 100 94 92
NOsize 128 82 49 53
sim 0.021 0.009 0.069 0.060
C 0.026 0.163 0.074 0.126
𝜏 2.9 2.2 1.8 1.6
weightBias 0.00 0.05 0.10 0.10
errorCorrection 0.96 0.94 0.70 0.71
scoreWindowSize 2 2 2 2
Table 14 Hyperparameters on the modified NAB dataset for the SPIRIT model

1402
1 3
A.4 Anomaly detection results
See Tables 15, 16 and 17.
Table 15 Average F1 score to detect the anomaly types of the NAB dataset for one dimension
Anomaly SPIRIT OeSNN-A OeSNN-B OeSNN-C OeSNN-D
Signal dropouts – 0.01 0.64 0.00 0.86
Signal drift – 0.02 0.51 0.49 0.72
Lower amplitude – 0.61 0.66 0.00 0.96
Higher amplitude – 0.88 0.83 0.92 0.88
Cycle dropouts – 0.27 0.62 0.00 0.00
Single peaks – 0.09 0.00 0.18 0.11
Change of frequency – 0.15 0.46 0.00 0.04
Increase in noise – 0.00 0.00 0.31 0.20
Increase in amplitude – 0.00 0.49 0.48 0.47
Average 0.22 0.47 0.26 0.47
Table 16 Average F1 score to detect the anomaly types of the NAB dataset for two dimensions
Signal dropouts 0.70 0.01 0.61 0.72 0.77
Signal drift 0.34 0.02 0.50 0.75 0.74
Lower amplitude 0.46 0.60 0.61 0.96 0.93
Cycle dropouts 0.36 0.26 0.55 0.91 0.86
Single peaks 0.00 0.09 0.00 0.19 0.16
Average 0.42 0.22 0.44 0.61 0.61
Table 17 Average F1 score to detect the anomaly types of the NAB dataset for four dimensions
Signal dropouts 0.58 0.01 0.61 0.70 0.76
Signal drift 0.44 0.02 0.50 0.66 0.73
Lower amplitude 0.00 0.60 0.61 0.88 0.95
Cycle dropouts 0.00 0.26 0.55 0.87 0.89
Single peaks 0.00 0.09 0.00 0.14 0.19
Average 0.31 0.22 0.44 0.57 0.62

1403
1 3
Fig. 9 Visualization
of
the
outlier
score
of
the
OeSNN
models
and
the
SPIRIT
reference
algorithm
on
the
adapted
NAB
dataset
with
four
dimensions

1404
1 3
Fig. 10 Visualization
of
the
outlier
score
of
the
OeSNN
models
and
the
SPIRIT
reference
algorithm
on
the
adapted
NAB
dataset
with
four
dimensions

1405
1 3
A.5 Visualization of the outlier scores for different algorithms
See Figs. 9 and 10.
Author Contributions Dennis Bäßler, Tobias Kortus and Gabriele Gühring contributed in an equal way to
this paper.
Funding Open Access funding enabled and organized by Projekt DEAL. No funding was received to assist
with the preparation of this manuscript.
Availability of data and material The data that support the findings of this study are available upon reason-
able request from the authors.
Code Availability The code that support the findings of this study are available upon reasonable request
from the authors.
Declarations
Conflict of interest The authors declare that they have no conflict of interest.
Ethics approval Not Applicable.
Consent for publication Upon acceptance we will either grant the Publisher an exclusive license to publish
the article or will transfer copyright of the article to the Publisher.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License,
which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long
as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Com-
mons licence, and indicate if changes were made. The images or other third party material in this article
are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons licence and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly
from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
References
Aggarwal, C. (2013). Outlier analysis. New York, NY: Springer-Verlag.
Ahmad, S., Lavin, A., Purdy, S., Agha, Z. (2017). Unsupervised real-time anomaly detection for
streaming data. Neurocomputing. https://doi.org/10.1016/j.neucom.2017.04.070
Akiba, T., Sano, S., Yanase, T., Ohta, T., Koyama, M. (2019). Optuna: A next-generation hyperparam-
eter optimization framework. In: Proceedings of the 25th ACM SIGKDD International Conference
on Knowledge Discovery and Data Mining, Association for Computing Machinery, New York, NY,
USA, KDD ’19, pp. 2623–2631, https://doi.org/10.1145/3292500.3330701.
Amirshahi, A., Hashemi, M. (2019). Ecg classification algorithm based on stdp and r-stdp neural net-
works for real-time monitoring on ultra low-power personal wearable devices. IEEE Transactions
on Biomedical Circuits and Systems, 13(6), 1483–1493. https://doi.org/10.1109/tbcas.2019.29489
20.
Bear, M., Seidler, L., Engel, A., Held, A., Connors, B., Hornung, C., et al. (2016). Neurowissenschaften: Ein
grundlegendes Lehrbuch für Biologie. Springer, Berlin Heidelberg: Medizin und Psychologie.
Behnel, S., Bradshaw, R., Citro, C., Dalcin, L., Seljebotn, D. S., Smith, K. (2011). Cython: The best of
both worlds. Computing in Science Engineering, 13(2), 31–39.
Bergstra, J., Bardenet, R., Bengio, Y., Kégl, B. (2011). Algorithms for hyper-parameter optimization. In
Proceedings of the 24th international conference on neural information processing systems, Curran
Associates Inc., Red Hook, NY, USA, NIPS’11, pp. 2546–2554.

1406
1 3
Bergstra, J., Yamins, D., Cox, DD. (2013). Making a science of model search: Hyperparameter optimi-
zation in hundreds of dimensions for vision architectures. In Proceedings of the 30th international
conference on international conference on machine learning, Volume 28, JMLR.org, ICML’13, pp.
I-115–I-123.
Bianco, A., Garcia Ben, M., Martínez, E., Yohai, V. (2001). Outlier detection in regression models with
arima errors using robust estimates. Journal of Forecasting,20. https://doi.org/10.1002/for.768.
Bing, Z., Meschede, C., Huang, K., Chen, G., Rohrbein, F., Akl, M., Knoll, A. (2018a). End to End
Learning of Spiking Neural Network Based on R-STDP for a Lane Keeping Vehicle. Proceedings—
IEEE international conference on robotics and automation pp. 4725–4732. https://doi.org/10.1109/
ICRA.2018.8460482.
Bing, Z., Meschede, C., Röhrbein, F., Huang, K., Knoll, A. C. (2018). A survey of robotics control based
on learning-inspired spiking neural networks. Frontiers in Neurorobotics, 12, 35. https://doi.org/10.
3389/fnbot.2018.00035.
Breunig, MM., Kriegel, HP., Ng, RT., Sander, J. (2000). Lof: Identifying density-based local outliers. In
Proceedings of the 2000 ACM SIGMOD international conference on management of data, association
for computing machinery, New York, NY, USA, SIGMOD ’00, pp. 93–104. https://doi.org/10.1145/
342009.335388, https://doi.org/10.1145/342009.335388.
Chalapathy, R., Chawla, S. (2019). Deep learning for anomaly detection: A survey. arXiv:1901.03407.
Däubener, S., Schmitt, S., Wang, H., Bäck, T., peter, krause. (2019). Large anomaly detection in univariate
time series: An empirical comparison of machine learning algorithms. In 19th Industrial conference on
data mining ICDM 2019, Unknown.
Demertzis, K., Iliadis, L., Spartalis, S. (2017). A spiking one-class anomaly detection framework for
cyber-security on industrial control systems. In G. Boracchi, L. Iliadis, C. Jayne, A. Likas (Eds.),
Engineering applications of neural networks (pp. 122–134). Cham: Springer International Publishing.
Demertzis, K., Iliadis, L., Bougoudis, I. (2019). Gryphon: A semi-supervised anomaly detection system
based on one-class evolving spiking neural network. Neural Computing and Applications. https://doi.
org/10.1007/s00521-019-04363-x.
Fu, X., Luo, H., Zhong, S., LIN L,. (2019). Aircraft engine fault detection based on grouped convo-
lutional denoising autoencoders. Chinese Journal of Aeronautics,32(2), 296–307. https://doi.org/10.
1016/j.cja.2018.12.011, http://www.sciencedirect.com/science/article/pii/S1000936119300238.
Geiger, A., Liu, D., Alnegheimish, S., Cuesta-Infante, A., Veeramachaneni, K. (2020). Tadgan: Time
series anomaly detection using generative adversarial networks. arXiv:2009.07769.
Gerstner, W., Kistler, W. M. (2002). Spiking neuron models: Single neurons, populations, plasticity. Cam-
bridge University Press. https://doi.org/10.1017/CBO9780511815706.
Goodfellow, I., Bengio, Y., Courville, A. (2016). Deep learning. MIT Press.
Gühring, G., Baum, C., Kleschew, A., Schmid, D. (2019). Anomalie-erkennung. atp magazin, 61, 66.
https://doi.org/10.17560/atp.v61i5.2380.
Hau, M., Tong, H. (1989). A practical method for outlier detection in autoregressive time series model-
ling. Stochastic Hydrology and Hydraulics, 3, 241–260. https://doi.org/10.1007/BF01543459.
Hopkins, M., García, G., Bogdan, P., Furber, S. (2018). Spiking neural networks for computer vision.
Interface Focus, 8, 20180007. https://doi.org/10.1098/rsfs.2018.0007.
Hossin, M., Sulaiman, M. (2015). A review on evaluation metrics for data classification evaluations.
International Journal of Data Mining Knowledge Management Process, 5, 01–11. https://doi.org/
10.5121/ijdkp.2015.5201.
Jeni, LA., Cohn, JF., De La Torre, F. (2013). Facing imbalanced data–recommendations for the use of
performance metrics. In 2013 Humaine association conference on affective computing and intelligent
interaction, pp. 245–251.
Kasabov, N. (2006). Evolving connectionist systems: The knowledge engineering approach. Berlin, Heidel-
berg: Springer-Verlag.
King, A. (2012). Online k-means clustering of nonstationary data. Prediction Protect Report. p. 11.
Kriegel, HP., Kröger, P., Schubert, E., Zimek, A. (2009). Loop: Local outlier probabilities. In Proceed-
ings of the 18th ACM conference on information and knowledge management, association for com-
puting machinery, New York, NY, USA, CIKM ’09, pp. 1649-1652. https://doi.org/10.1145/1645953.
1646195, https://doi.org/10.1145/1645953.1646195.
Lapicque, L. (1907). Recherches quantitatives sur l’excitation électrique des nerfs traitée comme une
polarisation. Journal de Physiologie et de Pathologie Générale, 9, 620–635. https://doi.org/10.1007/
s00422-007-0189-6.
Lobo, J. L., Lan̄a, I., Del Ser, J., Bilbao, MN., Kasbov, N. (2018). Evolving Spiking Neural Networks
for online learning over drifting data streams. Neural Networks,108, 1–19. https://doi.org/10.1016/j.
neunet.2018.07.014.

1407
1 3
Lobo, J. L., Javier, D. S., Albert, B., Nikola, K. (2020). Spiking Neural Networks and online learning: An
overview and perspectives. Neural Networks, 121, 88–100. https://doi.org/10.1016/j.neunet.2019.09.
004.
Li, D., Chen, D., Jin, B., Shi, L., Goh, J., Ng, S. K. (2019). Mad-gan: Multivariate anomaly detection
for time series data with generative adversarial networks. In I. V. Tetko, V. Kůrková, P. Karpov,
F. Theis (Eds.), Artificial neural networks and machine learning—ICANN 2019: Text and time
series (pp. 703–716). Cham: Springer International Publishing.
Li, Y., Lu, A., Wu, X., Yuan, S. (2019b). Dynamic anomaly detection using vector autoregressive
model. Springer International Publishing, pp. 600–611. https://doi.org/10.1007/978-3-030-16148-
4_46.
Lindeberg, T. (2013). A computational theory of visual receptive fields. Biological Cybernetics, 107(6),
589–635.
Maas, W. (1997). Networks of spiking neurons: The third generation of neural network models. Trans
Soc Comput Simul Int, 14(4), 1659–1671.
Macia̧g, P. S., Kryszkiewicz, M., Bembenik, R., Lobo, J. L., Ser, J. D. (2021). Unsupervised anomaly
detection in stream data with online evolving spiking neural networks. Neural Networks,139, 118–139.
Mahajan, M., Nimbhorkar, P., Varadarajan, K. (2012). The planar k-means problem is NP-hard. Theo-
retical Computer Science, Elsevier, 442, 13–21.
Malhotra, P., Ramakrishnan, A., Anand, G., Vig, L., Agarwal, P., Shroff, G. (2016). Lstm-based
encoder-decoder for multi-sensor anomaly detection. ArXivarXiv:1607.00148
Moayedi, H., Masnadi-Shirazi, M. (2008). Arima model for network traffic prediction and anomaly
detection. In 2008 International symposium on information technology, Vol. 4, pp. 1–6.
Munir, M., Siddiqui, S. A., Dengel, A., Ahmed, S. (2019). Deepant: A deep learning approach for
unsupervised anomaly detection in time series. IEEE Access, 7, 1991–2005.
Panuku, L. N., Sekhar, C. C. (2008). Region-based encoding method using multi-dimensional gauss-
ians for networks of spiking neurons. In M. Ishikawa, K. Doya, H. Miyamoto, T. Yamakawa
(Eds.), Neural information processing (pp. 73–82). Heidelberg: Springer, Berlin Heidelberg, Berlin.
Papadimitriou, S., Sun, J., Faloutsos, C. (2005). Streaming pattern discovery in multiple time-series.
In Proceedings of the 31st international conference on very large data bases, VLDB endowment,
VLDB ’05, pp. 697–708.
Pokrajac, D., Lazarevic, A., Latecki, L. J. (2007). Incremental local outlier detection for data streams.
In 2007 IEEE symposium on computational intelligence and data mining, pp. 504–515.
Schliebs, S., Kasabov, N. (2013). Evolving spiking neural networks: A survey. Evolving Systems,4.
https://doi.org/10.1007/s12530-013-9074-9.
Schuman, CD., Potok, TE., Patton, RM., Birdwell, JD., Dean, ME., Rose, GS., Plank, JS. (2017). A
survey of neuromorphic computing and neural networks in hardware. arXiv:1705.06963.
Shukla, R., Sengupta, S. (2020). Scalable and robust outlier detector using hierarchical clustering
and long short-term memory (lstm) neural network for the internet of things. Internet of Things, 9,
100167. https://doi.org/10.1016/j.iot.2020.100167.
Stinner, V. (2017). Pep 564 – add new time functions with nanosecond resolution. https://www.python.
org/dev/peps/pep-0564/. [Online; accessed 16 August 2020].
Taddei, A., Distante, G., Emdin, M., Pisani, P., Moody, G. B., Zeelenberg, C., Marchesi, C. (2000).
European st-t database. https://doi.org/10.13026/C2D59Z.
Thorpe, S., Gautrais, J. (1998). Rank order coding. Computational Neuroscience: Trends in Research
pp. 113–118. https://doi.org/10.1007/978-1-4615-4831-7_19.
Wang, J., Belatreche, A., Maguire, L. P., McGinnity, T. M. (2017). Spiketemp: An enhanced rank-
order-based learning approach for spiking neural networks with adaptive structure. IEEE Transac-
tions on Neural Networks and Learning Systems, 28(1), 30–43.
Watts, M. (2009). A decade of Kasabov’s evolving connectionist systems: A review. Systems, Man, and
Cybernetics, Part C: Applications and Reviews, IEEE Transactions on, 39, 253–269. https://doi.
org/10.1109/TSMCC.2008.2012254.
Welford, B. P. (1962). Note on a method for calculating corrected sums of squares and products. Techno-
metrics, 4(3), 419–420. https://doi.org/10.1080/00401706.1962.10490022.
Wu, J., Yılmaz, E., Zhang, M., Li, H., Tan, K. C. (2020). Deep spiking neural networks for large
vocabulary automatic speech recognition. Frontiers in Neuroscience, 14, 199. https://doi.org/10.
3389/fnins.2020.00199.
Wu, R., Keogh, E. (2021). Current time series anomaly detection benchmarks are flawed and are cre-
ating the illusion of progress. IEEE Transactions on Knowledge and Data Engineering. https://doi.
org/10.1109/TKDE.2021.3112126.

1408
1 3
Wysoski, S. G., Benuskova, L., Kasabov, N. (2006). On-line learning with structural adaptation in
a network of spiking neurons for visual pattern recognition. In S. D. Kollias, A. Stafylopatis, W.
Duch, E. Oja (Eds.), Artificial neural networks—ICANN 2006 (pp. 61–70). Heidelberg: Springer,
Berlin Heidelberg, Berlin.
Wysoski, S. G., Benuskova, L., Kasabov, N. (2008). Adaptive spiking neural networks for audiovisual
pattern recognition (pp. 406–415). Berlin, Heidelberg: Springer-Verlag.
Xing, L., Demertzis, K., Yang. (2019). Identifying data streams anomalies by evolving spiking restricted
boltzmann machines. Neural Computing and Applications,31, 1–15. https://doi.org/10.1007/
s00521-019-04288-5c.
Yu, T., Zhu, H. (2020). Hyper-parameter optimization: A review of algorithms and applications. arXiv:
2003.05689.
Zhang, G. (2003). Time series forecasting using a hybrid Arima and neural network model. Neurocomput-
ing, 50, 159–175. https://doi.org/10.1016/S0925-2312(01)00702-0.
Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.

Bäßler2022_Article_UnsupervisedAnomalyDetectionIn.pdf

More Related Content

Similar to Bäßler2022_Article_UnsupervisedAnomalyDetectionIn.pdf (20)

More from TadiyosHailemichael (11)

Recently uploaded (20)

Bäßler2022_Article_UnsupervisedAnomalyDetectionIn.pdf