How NoSQL Fundamentally
Changed Machine Learning
2015/8/7
棚橋 耕太郎http://data-magnum.com/how-nosql-has-
fundamentally-changed-machine-learning/
1. predictive models
主なデータ:数値
主な手法:クラスタリング,回帰,分類器
rise of date lakes
NoSQLによりXML,JSONなどが利用可能に
2. recommenders
大規模データを用いてレコメンドを行う
ビジネス的に重要な領域
3. NLP
interpret general trends and sentiment
NLPは予測モデルにも使える
4. IOT
大量のデータはNoSQLによって処理される
5. image processing
画像処理やDLはまだ未熟な領域
Question answering on the
Facebook bAbi dataset using
recurrent neural networks and
175 lines of Python + Keras
http://smerity.com/articles/2015/keras_qa.html
question answering
Facebook bAbi tasks
人口の質問データセット
ロボットの評価に用いる
Why a synthetic dataset?
The vocabulary (set of words) is constrained, the sentences are
always well structured (the only noise is the noise we want to
challenge the model with), and the performance on specific tasks can
be tested without other tasks interferring
How do we approach the problem?
Word vectors
https://en.wikipedia.org/wiki/Recurrent_neural_network
performance
the results are comparable (and
occasionally superior) to those for
the LSTM baseline provided in
Weston et al.'s
comparison to FB baseline
まとめ
・question answering の機械がこんな簡単に作れる!
・いろいろ応用できそう(LINEのログを使いたい)
・theanoのような計算フレームワークの必要性を再認識

More Related Content

PPTX
ジャストシステムの形態素解析技術
PPTX
形態素解析器 売ってみた
PPTX
ジャストシステムの形態素解析技術 その2 機械学習編
PPTX
[DL輪読会]Deep Face Recognition: A Survey
PPTX
[DL輪読会]Learning Latent Dynamics for Planning from Pixels
PDF
【CVPR 2020 メタサーベイ】Scene Analysis and Understanding
PDF
効率的学習 / Efficient Training(メタサーベイ)
PPTX
【論文紹介】How Powerful are Graph Neural Networks?
ジャストシステムの形態素解析技術
形態素解析器 売ってみた
ジャストシステムの形態素解析技術 その2 機械学習編
[DL輪読会]Deep Face Recognition: A Survey
[DL輪読会]Learning Latent Dynamics for Planning from Pixels
【CVPR 2020 メタサーベイ】Scene Analysis and Understanding
効率的学習 / Efficient Training(メタサーベイ)
【論文紹介】How Powerful are Graph Neural Networks?

What's hot (20)

PDF
Tree-to-Sequence Attentional Neural Machine Translation (ACL 2016)
PDF
【メタサーベイ】Face, Gesture, and Body Pose
PPTX
20190509 gnn public
PDF
【CVPR 2019】Second-order Attention Network for Single Image Super-Resolution
PDF
アプリケーション開発における暗号化
PDF
Capsule Graph Neural Network
PDF
[DL輪読会]Training RNNs as Fast as CNNs
PDF
JMAT Groonga Tokenizer Talks
PPTX
Invariant Information Clustering for Unsupervised Image Classification and Se...
PPTX
Generating Better Search Engine Text Advertisements with Deep Reinforcement L...
PPTX
[DL輪読会]Dream to Control: Learning Behaviors by Latent Imagination
PPTX
Playing Atari with Six Neurons
PDF
A PID Controller Approach for Stochastic Optimization of Deep Networks
PDF
End-to-end Recovery of Human Shape and Pose
PDF
NIPS2013読み会: More Effective Distributed ML via a Stale Synchronous Parallel P...
PDF
Graph Attention Network
PPTX
Graph U-Nets
PDF
[DLHacks 実装] DeepPose: Human Pose Estimation via Deep Neural Networks
PDF
論文 Solo Advent Calendar
PDF
[DL輪読会] Spectral Norm Regularization for Improving the Generalizability of De...
Tree-to-Sequence Attentional Neural Machine Translation (ACL 2016)
【メタサーベイ】Face, Gesture, and Body Pose
20190509 gnn public
【CVPR 2019】Second-order Attention Network for Single Image Super-Resolution
アプリケーション開発における暗号化
Capsule Graph Neural Network
[DL輪読会]Training RNNs as Fast as CNNs
JMAT Groonga Tokenizer Talks
Invariant Information Clustering for Unsupervised Image Classification and Se...
Generating Better Search Engine Text Advertisements with Deep Reinforcement L...
[DL輪読会]Dream to Control: Learning Behaviors by Latent Imagination
Playing Atari with Six Neurons
A PID Controller Approach for Stochastic Optimization of Deep Networks
End-to-end Recovery of Human Shape and Pose
NIPS2013読み会: More Effective Distributed ML via a Stale Synchronous Parallel P...
Graph Attention Network
Graph U-Nets
[DLHacks 実装] DeepPose: Human Pose Estimation via Deep Neural Networks
論文 Solo Advent Calendar
[DL輪読会] Spectral Norm Regularization for Improving the Generalizability of De...
Ad

Viewers also liked (9)

PDF
Cythonの一喜一憂
PDF
変数の入れ替え(SWAPPING)で最速の方法は?
PDF
SWARでpop countをしよう
PDF
2.2. map reduce and the new software stack
PDF
deep learning library coyoteの開発(CNN編)
PDF
DSP開発におけるSpark MLlibの活用
PDF
NIPS Paper Reading, Data Programing
PDF
変分推論法(変分ベイズ法)(PRML第10章)
PDF
WSDM2016読み会 Collaborative Denoising Auto-Encoders for Top-N Recommender Systems
Cythonの一喜一憂
変数の入れ替え(SWAPPING)で最速の方法は?
SWARでpop countをしよう
2.2. map reduce and the new software stack
deep learning library coyoteの開発(CNN編)
DSP開発におけるSpark MLlibの活用
NIPS Paper Reading, Data Programing
変分推論法(変分ベイズ法)(PRML第10章)
WSDM2016読み会 Collaborative Denoising Auto-Encoders for Top-N Recommender Systems
Ad

How nosql fundamentally changed machine learning?