Sequence to sequence (S2S) modeling for time series forecasting

Arun Kejariwal (Independent), Ira Cohen (Anodot)

11:55–12:35 Wednesday, 16 October 2019

Location: King's Suite - Sandringham

Models and Methods

Secondary topics: Deep Learning, Machine Learning, Temporal data and time-series

Average rating:

(4.00, 5 ratings)

Who is this presentation for?

Practitioners in the industry

Level

Intermediate

Description

S2S modeling using neural networks has been increasingly becoming mainstream in the recent years. In particular, it’s been used for applications such as, but not limited to, speech recognition, language translation, and question answering. More recently, S2S has also been used for applications based on time series data. Specifically, S2S modeling-based real-time anomaly detection and forecasting is being actively explored.

Arun Kejariwal and Ira Cohen provide an overview of S2S and early use cases of S2S. They walk you through how S2S modeling can be leveraged for the aforementioned use cases, visualization, real-time anomaly detection, and forecasting. You’ll learn how a multilayered long short-term memory (LSTM) is used encode to the input time series and then a deep LSTM is used to decode. In case of anomaly detection, the output is married with “traditional” statistical approaches for anomaly detection. Conceivably, any of the many variants of LSTM or recurrent neural network (RNN) alternatives of LSTM can be used to trade-off accuracy and speed. Further, given that LSTMs operate sequentially and are quite slow to train, Arun and Ira shed light on how architectures such as, but not limited to, convolutional neural network (CNN) and self-attention network (SAN) can be leveraged to achieve significant improvements in accuracy.

You’ll get a look at a concrete case study to illustrate the use of S2S for both real-time anomaly detection and forecasting for time series data.

What you'll learn

Learn how to use S2S for both real-time anomaly detection and forecasting for time series data

Arun Kejariwal

Independent

Arun Kejariwal is an independent lead engineer. Previously, he was he was a statistical learning principal at Machine Zone (MZ), where he led a team of top-tier researchers and worked on research and development of novel techniques for install-and-click fraud detection and assessing the efficacy of TV campaigns and optimization of marketing campaigns, and his team built novel methods for bot detection, intrusion detection, and real-time anomaly detection; and he developed and open-sourced techniques for anomaly detection and breakout detection at Twitter. His research includes the development of practical and statistically rigorous techniques and methodologies to deliver high performance, availability, and scalability in large-scale distributed clusters. Some of the techniques he helped develop have been presented at international conferences and published in peer-reviewed journals.

Ira Cohen

Anodot

Ira Cohen is a cofounder and chief data scientist at Anodot, where he’s responsible for developing and inventing the company’s real-time multivariate anomaly detection algorithms that work with millions of time series signals. He holds a PhD in machine learning from the University of Illinois at Urbana-Champaign and has over 12 years of industry experience.