Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Seq2seq Model on Time-series Data: Training and Serving with TensorFlow - Masood Krohy

Seq2seq Model on Time-series Data: Training and Serving with TensorFlow - Masood Krohy

Masood Krohy at April 9, 2019 event of montrealml.dev

Title: Seq2seq Model on Time-series Data: Training and Serving with TensorFlow

Presentation/Demo video: https://www.youtube.com/watch?v=H6Sv0jkwIl8

Summary: Seq2seq models are a class of Deep Learning models that have provided state-of-the-art solutions to language problems recently. They also perform very well on numerical, time-series data which is of particular interest in finance and IoT, among others. In this hands-on demo/code walkthrough, we explain the model development and optimization with TensorFlow (its low-level API). We then serve the model with TensorFlow Serving and show how to write a client to communicate with TF Serving over the network and use/plot the received predictions.

Code on GitHub: https://github.com/patternedscience/time-series-tf-serving

Bio: Masood Krohy is a Data Science Platform Architect/Advisor and most recently acted as the Chief Architect of UniAnalytica, an advanced data science platform with wide, out-of-the-box support for time-series and geospatial use cases. He has worked with several corporations in different industries in the past few years to design, implement and productionize Deep Learning and Big Data products. He holds a Ph.D. in computer engineering.

PatternedScience

April 09, 2019
Tweet

More Decks by PatternedScience

Other Decks in Programming

Transcript

  1. Copyright © 2019, PatternedScience Inc.
    www.patterned.science
    Seq2seq Model on Time-series Data:
    Training and Serving with TensorFlow
    Presenter
    Masood Krohy, Ph.D.
    Version 1.0
    April 9, 2019

    View Slide

  2. 2
    Copyright © 2019, PatternedScience Inc.
    Final Notes /
    Q&A
    Model serving
    and sample
    client code
    Model training,
    optimization,
    exporting
    Presenter bio
    Presentation Layout

    View Slide

  3. 3
    Copyright © 2019, PatternedScience Inc.
    Ph.D. in Computer Engineering
    Analytical modeling of botnets. Validated by data collected in industry. 3 top publications.
    Senior Analyst, Rogers
    Managing the analytics reporting/statistical analyses of the national benchmarking program.
    Data Scientist, Intact
    First Data Scientist of the company. Led the Big Data mining project for the UBI program.
    Lead Data Scientist, CN
    Implemented an object-within-object detection system to detect cracks in railway equipment.
    Masood Krohy
    Presenter Bio
    2013
    Sr Data Science Advisor, B.Yond
    Implemented a pattern detection system for stream of alarms coming from telecom devices.
    Chief Architect, UniAnalytica (advanced data science platform)
    Platform contains the open-source technologies presented here, among many others.
    2014
    2016
    2017
    2018
    2019
    Data Science Platform Architect & Advisor

    View Slide

  4. Model training,
    optimization,
    exporting

    View Slide

  5. 5
    Copyright © 2019, PatternedScience Inc.
    Notes on model optimization
    A better (visual!) analysis of grid search results with Parallel Coordinates visualization will be
    introduced in a later presentation.
    A better analysis of grid search results
    01
    ● When is it needed? when optimizing over several parameters simultaneously.
    A naïve grid search would need an astronomically large number of experiments.
    ● A more advanced hyperparam search with HyperBand, which outperforms grid search,
    random search and Bayesian optimization, will also be introduced in a later presentation.
    We can do better than grid search (if needed)
    02

    View Slide

  6. 6
    Copyright © 2019, PatternedScience Inc.
    2. Spark & TensorFlow/scikit-learn
    Distributed grid search with Spark and
    TensorFlow/scikit-learn (small datasets,
    perfectly parallel)
    5. Interpretable AI
    Images - Classification with visual explanation for
    classifications using Class Activation Maps
    3. Ray Tune & TensorFlow/scikit-learn
    Intelligent, distributed hyperparam search with
    Asynchronous Hyperband, Ray Tune, and
    TensorFlow/scikit-learn (small datasets,
    perfectly parallel)
    4. ML on images
    Images - TensorFlow Object Detection API (intro)
    1. Horovod & TensorFlow
    Distributed Deep Learning with
    TensorFlow and Horovod (large datasets,
    data parallelism)
    UniAnalytica Platform
    Machine Learning Stack
    www.unianalytica.com

    View Slide

  7. Model serving
    and sample
    client code

    View Slide

  8. 8
    Copyright © 2019, PatternedScience Inc.
    Final Notes
    On top of what was shown, using
    other techniques including protobuf
    for data serialization and gRPC for
    communication can lead to better
    performance and scalability (if this is
    needed).
    Optimizing client code
    https://github.com/patternedscience/time-series-tf-serving
    Code and results are on GitHub
    To discuss technical results and give feedback, please email
    us at [email protected]
    We love feedback!
    https://www.linkedin.com/company/patterned-science/
    Follow us for future results & announcements!

    View Slide

  9. Q&A

    View Slide