SlideShare a Scribd company logo
1
Imputing Missing Events in Continuous-Time Event
Streams (ICML 2019)
Akitoshi Kimura, Taniguchi Lab, Waseda University
書誌情報
• 著者:
– Hongyuan Mei
• Department of Computer Science, Johns Hopkins University, USA
– Guanghui Qin
• Department of Physics, Peking University, China
– Jason Eisner
• Department of Computer Science, Johns Hopkins University, USA
• 学会:
– ICML 2019 Oral, Poster
2
概要
• イベント系列を Neural Hawkes process でモデリング
• イベント系列の観測されなかった部分の補完
– Medical records, Competitive games, User interface interactions
• 提案分布に bidirectional continuous-time LSTM を適用
3
背景: neural Hawkes process
• Mei & Eisner(2017)
• 互いに影響しあうイベントのモデル
4
既存手法の問題点
• 事後分布 𝑝𝑝 𝑧𝑧 𝑥𝑥 を求めるのは難しい
• 𝑥𝑥: neural Hawkes process から得られた(不完全な)な観測データ
• 𝑧𝑧: (生成されたが)観測されなかったデータ
– Hawkes process でも MCMC の必要
– Efficient transition kernel が必要
– Neural Hawkes process ではできない
5
提案手法
• 一般的な sequential Monte Carlo で事後分布からサンプル
• Particle filtering
– 各時点では、過去の観測データ、非観測データを考慮に入れる
• Particle smoothing
– 各時点では、さらに将来の観測データも考慮に入れる
6
Particle filtering
• タクシーへの乗車下車の例
7
Particle smoothing
• 将来の観測データも活用
8
モデル
• 𝑝𝑝𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚: neural Hawkes process
– Intensity function:
– History:
– Hidden state vector at time t:
• 𝑝𝑝𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚: missing at random: 𝑧𝑧 に依存しない
– missing not at random: 𝑧𝑧 に依存する
9
Sequential Monte Carlo
• 𝑝𝑝 𝑧𝑧 𝑥𝑥 からサンプリングは難しい
• 重点サンプリングを用いる
– 𝑞𝑞 𝑧𝑧 𝑥𝑥 : 提案分布からサンプリング
– ⁄𝑝𝑝 𝑧𝑧 𝑥𝑥 𝑞𝑞 𝑧𝑧 𝑥𝑥 に比例する重みをつける
• Ensemble of weighted particles:
– Importance weights:
10
Particle filtering and particle smoothing
• Particle filtering
– Intensity function:
– History:
• All observed and unobserved events
• Particle smoothing
– Intensity function:
– Future:
• All observed events that happen after 𝑡𝑡
11
提案分布の学習
• 𝑝𝑝 𝑧𝑧 𝑥𝑥 を近似するために KL divergence について最適化
• Linearly combined divergence
– Gradient of inclusive KL divergence
– Gradient of exclusive KL divergence
12
損失関数
• Minimum Bayes Risk decoding, consensus decoding
– Optimal transport distance:
– The set of all alignments between 𝑧𝑧 and 𝑧𝑧∗
:
– The total cost given the alignment 𝑎𝑎:
• decomposed as
13
実験
• Missing data mechanisms
• Datasets
– Synthetic datasets
– Elevator system dataset
– New York city taxi dataset
14
手順
• Training data から 𝑝𝑝𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 を学習( 𝑝𝑝𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 は既知とする)
• 𝑝𝑝 𝑧𝑧 𝑥𝑥 を近似するように 𝑞𝑞 𝑧𝑧 𝑥𝑥 を最適化
• 𝑞𝑞 𝑧𝑧 𝑥𝑥 から weighted particle
をサンプリング
• ̂𝑧𝑧: consensus sequence を得る
• Optimal transport distance L ̂𝑧𝑧, 𝑧𝑧∗
を評価して比較
– Particle filtering
– Particle smoothing
15
Data fitting results
• Scatterplots of neural Hawkes particle smoothing (yaxis)
vs. particle filtering (x-axis)
16
Decoding results
• Optimal transport distance of particle smoothing (red
triangle) vs. particle filtering (blue circle)
17
結論
• bidirectional recurrent neural network によるイベント系列
の予測は初
• イベント系列どうしを評価する optimal transport distance
を提案
• Consensus sequence を得る方法を与えた
• 提案手法は人工データでも実データでも非観測系列の推測に
おいて効果的であった 18
参考文献
• Mei, H. and Eisner, J. The neural Hawkes process: A
neurally self-modulating multivariate point process. In
Advances in Neural Information Processing Systems
(NIPS), 2017.
19

More Related Content

PPTX
[DL輪読会]Learning convolutional neural networks for graphs
PPTX
[DL輪読会]Temporal DifferenceVariationalAuto-Encoder
PPTX
NeurIPS2019参加報告
PPTX
【DL輪読会】時系列予測 Transfomers の精度向上手法
PDF
ICML2017 参加報告会 山本康生
PDF
[DL輪読会]The Neural Process Family−Neural Processes関連の実装を読んで動かしてみる−
PDF
[DL輪読会]Z-Forcing: Training Stochastic Recurrent Networks (NIPS2017)
PDF
[DL輪読会]Convolutional Conditional Neural Processesと Neural Processes Familyの紹介
[DL輪読会]Learning convolutional neural networks for graphs
[DL輪読会]Temporal DifferenceVariationalAuto-Encoder
NeurIPS2019参加報告
【DL輪読会】時系列予測 Transfomers の精度向上手法
ICML2017 参加報告会 山本康生
[DL輪読会]The Neural Process Family−Neural Processes関連の実装を読んで動かしてみる−
[DL輪読会]Z-Forcing: Training Stochastic Recurrent Networks (NIPS2017)
[DL輪読会]Convolutional Conditional Neural Processesと Neural Processes Familyの紹介

Similar to [DL輪読会]Imputing Missing Events in Continuous-Time Event Streams (ICML 2019) (20)

PDF
第61回CV勉強会「CVPR2024読み会」(前編)発表資料:State Space Models for Event Cameras
PDF
[DL Hacks] Deterministic Variational Inference for RobustBayesian Neural Netw...
PDF
Large Scale Incremental Learning
PDF
第2回nips+読み会: Learning to learn by gradient decent by gradient decent
PDF
MLaPP 24章 「マルコフ連鎖モンテカルロ法 (MCMC) による推論」
PPTX
[DL輪読会] Learning Finite State Representations of Recurrent Policy Networks (I...
PPTX
Active Learning と Bayesian Neural Network
PDF
卒業論文「主張と根拠のクラスタを用いた 多様な主張を提示するニュース推薦手法の提案」
PPTX
DNNの曖昧性に関する研究動向
PPTX
【DL輪読会】AdaptDiffuser: Diffusion Models as Adaptive Self-evolving Planners
PDF
異常行動検出入門 – 行動データ時系列のデータマイニング –
PDF
Dst_prediction_202203_TECH_in_Kyoto.pdf
PDF
Neural networks for Graph Data NeurIPS2018読み会@PFN
PDF
偏りのある時系列データ の分類について
PDF
Journal club dec24 2015 splice site prediction using artificial neural netw...
PDF
SSII2022 [OS3-02] Federated Learningの基礎と応用
PDF
【DL輪読会】Can Neural Network Memorization Be Localized?
PDF
Deep Learningの基礎と応用
PDF
201803NC
PDF
Targeting EEG/LFP Synchrony with Neural Nets
第61回CV勉強会「CVPR2024読み会」(前編)発表資料:State Space Models for Event Cameras
[DL Hacks] Deterministic Variational Inference for RobustBayesian Neural Netw...
Large Scale Incremental Learning
第2回nips+読み会: Learning to learn by gradient decent by gradient decent
MLaPP 24章 「マルコフ連鎖モンテカルロ法 (MCMC) による推論」
[DL輪読会] Learning Finite State Representations of Recurrent Policy Networks (I...
Active Learning と Bayesian Neural Network
卒業論文「主張と根拠のクラスタを用いた 多様な主張を提示するニュース推薦手法の提案」
DNNの曖昧性に関する研究動向
【DL輪読会】AdaptDiffuser: Diffusion Models as Adaptive Self-evolving Planners
異常行動検出入門 – 行動データ時系列のデータマイニング –
Dst_prediction_202203_TECH_in_Kyoto.pdf
Neural networks for Graph Data NeurIPS2018読み会@PFN
偏りのある時系列データ の分類について
Journal club dec24 2015 splice site prediction using artificial neural netw...
SSII2022 [OS3-02] Federated Learningの基礎と応用
【DL輪読会】Can Neural Network Memorization Be Localized?
Deep Learningの基礎と応用
201803NC
Targeting EEG/LFP Synchrony with Neural Nets
Ad

More from Deep Learning JP (20)

PPTX
【DL輪読会】事前学習用データセットについて
PPTX
【DL輪読会】 "Learning to render novel views from wide-baseline stereo pairs." CVP...
PPTX
【DL輪読会】Zero-Shot Dual-Lens Super-Resolution
PPTX
【DL輪読会】BloombergGPT: A Large Language Model for Finance arxiv
PPTX
【DL輪読会】マルチモーダル LLM
PDF
【 DL輪読会】ToolLLM: Facilitating Large Language Models to Master 16000+ Real-wo...
PPTX
【DL輪読会】AnyLoc: Towards Universal Visual Place Recognition
PPTX
【DL輪読会】Hopfield network 関連研究について
PPTX
【DL輪読会】SimPer: Simple self-supervised learning of periodic targets( ICLR 2023 )
PDF
【DL輪読会】RLCD: Reinforcement Learning from Contrast Distillation for Language M...
PDF
【DL輪読会】"Secrets of RLHF in Large Language Models Part I: PPO"
PPTX
【DL輪読会】"Language Instructed Reinforcement Learning for Human-AI Coordination "
PPTX
【DL輪読会】Llama 2: Open Foundation and Fine-Tuned Chat Models
PDF
【DL輪読会】"Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware"
PPTX
【DL輪読会】Parameter is Not All You Need:Starting from Non-Parametric Networks fo...
PDF
【DL輪読会】Drag Your GAN: Interactive Point-based Manipulation on the Generative ...
PDF
【DL輪読会】Self-Supervised Learning from Images with a Joint-Embedding Predictive...
PPTX
【DL輪読会】Towards Understanding Ensemble, Knowledge Distillation and Self-Distil...
PPTX
【DL輪読会】VIP: Towards Universal Visual Reward and Representation via Value-Impl...
PDF
【DL輪読会】Deep Transformers without Shortcuts: Modifying Self-attention for Fait...
【DL輪読会】事前学習用データセットについて
【DL輪読会】 "Learning to render novel views from wide-baseline stereo pairs." CVP...
【DL輪読会】Zero-Shot Dual-Lens Super-Resolution
【DL輪読会】BloombergGPT: A Large Language Model for Finance arxiv
【DL輪読会】マルチモーダル LLM
【 DL輪読会】ToolLLM: Facilitating Large Language Models to Master 16000+ Real-wo...
【DL輪読会】AnyLoc: Towards Universal Visual Place Recognition
【DL輪読会】Hopfield network 関連研究について
【DL輪読会】SimPer: Simple self-supervised learning of periodic targets( ICLR 2023 )
【DL輪読会】RLCD: Reinforcement Learning from Contrast Distillation for Language M...
【DL輪読会】"Secrets of RLHF in Large Language Models Part I: PPO"
【DL輪読会】"Language Instructed Reinforcement Learning for Human-AI Coordination "
【DL輪読会】Llama 2: Open Foundation and Fine-Tuned Chat Models
【DL輪読会】"Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware"
【DL輪読会】Parameter is Not All You Need:Starting from Non-Parametric Networks fo...
【DL輪読会】Drag Your GAN: Interactive Point-based Manipulation on the Generative ...
【DL輪読会】Self-Supervised Learning from Images with a Joint-Embedding Predictive...
【DL輪読会】Towards Understanding Ensemble, Knowledge Distillation and Self-Distil...
【DL輪読会】VIP: Towards Universal Visual Reward and Representation via Value-Impl...
【DL輪読会】Deep Transformers without Shortcuts: Modifying Self-attention for Fait...
Ad

[DL輪読会]Imputing Missing Events in Continuous-Time Event Streams (ICML 2019)

  • 1. 1 Imputing Missing Events in Continuous-Time Event Streams (ICML 2019) Akitoshi Kimura, Taniguchi Lab, Waseda University
  • 2. 書誌情報 • 著者: – Hongyuan Mei • Department of Computer Science, Johns Hopkins University, USA – Guanghui Qin • Department of Physics, Peking University, China – Jason Eisner • Department of Computer Science, Johns Hopkins University, USA • 学会: – ICML 2019 Oral, Poster 2
  • 3. 概要 • イベント系列を Neural Hawkes process でモデリング • イベント系列の観測されなかった部分の補完 – Medical records, Competitive games, User interface interactions • 提案分布に bidirectional continuous-time LSTM を適用 3
  • 4. 背景: neural Hawkes process • Mei & Eisner(2017) • 互いに影響しあうイベントのモデル 4
  • 5. 既存手法の問題点 • 事後分布 𝑝𝑝 𝑧𝑧 𝑥𝑥 を求めるのは難しい • 𝑥𝑥: neural Hawkes process から得られた(不完全な)な観測データ • 𝑧𝑧: (生成されたが)観測されなかったデータ – Hawkes process でも MCMC の必要 – Efficient transition kernel が必要 – Neural Hawkes process ではできない 5
  • 6. 提案手法 • 一般的な sequential Monte Carlo で事後分布からサンプル • Particle filtering – 各時点では、過去の観測データ、非観測データを考慮に入れる • Particle smoothing – 各時点では、さらに将来の観測データも考慮に入れる 6
  • 9. モデル • 𝑝𝑝𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚: neural Hawkes process – Intensity function: – History: – Hidden state vector at time t: • 𝑝𝑝𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚: missing at random: 𝑧𝑧 に依存しない – missing not at random: 𝑧𝑧 に依存する 9
  • 10. Sequential Monte Carlo • 𝑝𝑝 𝑧𝑧 𝑥𝑥 からサンプリングは難しい • 重点サンプリングを用いる – 𝑞𝑞 𝑧𝑧 𝑥𝑥 : 提案分布からサンプリング – ⁄𝑝𝑝 𝑧𝑧 𝑥𝑥 𝑞𝑞 𝑧𝑧 𝑥𝑥 に比例する重みをつける • Ensemble of weighted particles: – Importance weights: 10
  • 11. Particle filtering and particle smoothing • Particle filtering – Intensity function: – History: • All observed and unobserved events • Particle smoothing – Intensity function: – Future: • All observed events that happen after 𝑡𝑡 11
  • 12. 提案分布の学習 • 𝑝𝑝 𝑧𝑧 𝑥𝑥 を近似するために KL divergence について最適化 • Linearly combined divergence – Gradient of inclusive KL divergence – Gradient of exclusive KL divergence 12
  • 13. 損失関数 • Minimum Bayes Risk decoding, consensus decoding – Optimal transport distance: – The set of all alignments between 𝑧𝑧 and 𝑧𝑧∗ : – The total cost given the alignment 𝑎𝑎: • decomposed as 13
  • 14. 実験 • Missing data mechanisms • Datasets – Synthetic datasets – Elevator system dataset – New York city taxi dataset 14
  • 15. 手順 • Training data から 𝑝𝑝𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 を学習( 𝑝𝑝𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 は既知とする) • 𝑝𝑝 𝑧𝑧 𝑥𝑥 を近似するように 𝑞𝑞 𝑧𝑧 𝑥𝑥 を最適化 • 𝑞𝑞 𝑧𝑧 𝑥𝑥 から weighted particle をサンプリング • ̂𝑧𝑧: consensus sequence を得る • Optimal transport distance L ̂𝑧𝑧, 𝑧𝑧∗ を評価して比較 – Particle filtering – Particle smoothing 15
  • 16. Data fitting results • Scatterplots of neural Hawkes particle smoothing (yaxis) vs. particle filtering (x-axis) 16
  • 17. Decoding results • Optimal transport distance of particle smoothing (red triangle) vs. particle filtering (blue circle) 17
  • 18. 結論 • bidirectional recurrent neural network によるイベント系列 の予測は初 • イベント系列どうしを評価する optimal transport distance を提案 • Consensus sequence を得る方法を与えた • 提案手法は人工データでも実データでも非観測系列の推測に おいて効果的であった 18
  • 19. 参考文献 • Mei, H. and Eisner, J. The neural Hawkes process: A neurally self-modulating multivariate point process. In Advances in Neural Information Processing Systems (NIPS), 2017. 19