TensorFlow for Building Deep Learning networks

Mar 25 @ 09:00 AM - 02:00 PM

NYC Seminar and Conference Center, NYC

Brought to you by

TensorFlow/Keras is becoming the industry standard for machine learning. After several successful training events, Alex Dimakis, Ben Lau and Nick Vasiloglou are presenting a one day event on advanced Machine Learning and TensorFlow. The event will be divided into two parallel tracks. All you need to know is python. We will help you learn the rest!

In the first track the participants will get exposed to the most influential deep learning architecture the Generative Adversarial Networks (GANs). GANs can synthesize images from noise or sketches. Along with LSTMs that can parse and generate text can generate art and poetry. The LSTM technology has also shown very good performance in modelling complex time series such as financial data.

In the second track the participants will learn the fundamentals of Deep Reinforcement Learning the basic ingredient for building AI agents and bots. One of the most intriguing application is agents for playing computer games. In this track you will learn how to code an agent in Keras that can drive a virtual car in the TORCS game environment.

Participants will have access to the complete material from both tracks. This event is sponsored by Metis.

Participants are expected to have TensorFlow 1.0, NumPy and Pandas installed on their laptops. Prior experience with TensorFlow is not required.

Schedule

  • Track 1
  • Track 2

Track 1

-

Deep Learning

  • Convolutional Networks
  • Generative Adversarial Networks
  • Word embeddings for language processing
  • Transfer Learning
  • Recurrent Neural Networks and LSTMs
-

Working with Images

  • Understanding and using Generative Models
  • Classifying images
-

Working with Time-Series and Text

  • Word2Vec
  • LSTMs for time-series and text modeling.
  • Text classification
  • Using TensorFlow and Keras to code these models

Track 2

-

Building AI agents and bots that play computer games. (Reinforcement Learning)

  • Overview
  • Introduction to RL : Value iteration vs Policy iteration
  • Reinforcement learning via Deep Q Networks (Theory)
  • Reinforcement learning via Deep Deterministic Policy Gradient (Theory)
  • Advanced Topic: Replay buffer/target networks, double-Q learning, A3C model
  • Playing DOOM using Deep Q Networks
  • Playing TORCS using Deep Deterministic Policy Gradient
  • Coding a game in OpenAI GYM

Location

NYC Seminar and Conference Center71 West 23rd StreetNew York, NY 10010