O'Reilly logo
live online training icon Live Online training

Building a Deep Learning Model using Tensorflow

Build a production ready deep learning pipeline with Keras using robust architectures and optimization frameworks

Rudy Lai

Deep learning is an exciting topic, and Tensorflow, Google’s open source deep learning framework is rapidly maturing. Developers are flooded with choice when it comes to tutorials around Tensorflow, but there hasn’t been an end-to-end course that shows you how to create production ready applications powered by deep learning.

In “Building a Deep Learning Model using TensorFlow and Keras”, we offer a course that brings you through the process of building a real world deep learning system. Using Bitcoin market price data as a dataset, we step through data cleaning, model architecture search, evaluation and hyperparameter optimization, and ending with creating an API using Flask. If you are a data science hobbyist or developer, this course will be perfect to bring your deep learning skills to the next level and into the real world.

What you'll learn-and how you can apply it

  • Understand how to analyze and clean data to prepare any dataset for machine learning
  • Discover the tradeoffs between different architectures to leverage optimal biases
  • Use Keras to quickly implement new deep learning ideas
  • Evaluate deep learning models to pick the optimal set of weights and hyperparameters
  • Use Tensorboard to visualize your training and understand your network

This training course is for you because...

You are interested in a practical, hands on guide to the three components of a deep learning application: data, model, and infrastructure. We spend equal amounts of time on each pillar to ensure that participants have a clear idea of how to tackle the issues of each stage. By combining clear presentations and clean code, this course focuses on teaching you how to create a robust deep learning pipeline, so that you can apply deep learning to actual problems rather than toy examples.

Prerequisites

  • Familiarity with the following is a must:
  • Python 3
  • Basics of neural networks
  • Knowledge of statistics, probability, linear algebra, and calculus
  • Experience with data-provisioning systems, including filesystems (local and remote) and databases (SQL and NoSQL)

Recommended Preparation:

Materials, downloads, or Supplemental Content needed in advance:

  • Visual Studio Code
  • Python 3
  • TensorFlow 1.4
  • Keras

Installation and Setup:

Course Materials:

About your instructor

  • Rudy Lai is the founder of QuantCopy, a sales acceleration startup using AI to write sales emails to prospects. By taking in leads from your pipelines, QuantCopy researches them online and generates sales emails from that data. It also has a suite of email automation tools to schedule, send, and track email performance - key analytics that all feedback into how our AI generates content.

    Prior to founding QuantCopy, Rudy ran HighDimension.IO, a machine learning consultancy, where he experienced firsthand the frustrations of outbound sales and prospecting. As a founding partner, he helped startups and enterprises with HighDimension.IO’s Machine-Learning-as-a-Service, allowing them to scale up data expertise in the blink of an eye.

    In the first part of his career, Rudy spent 5+ years in quantitative trading at leading investment banks such as Morgan Stanley. This valuable experience allowed him to witness the power of data, but also the pitfalls of automation using data science and machine learning. Quantitative trading was also a great platform to learn deeply about reinforcement learning and supervised learning topics in a commercial setting.

    Rudy holds a Computer Science degree from Imperial College London, where he was part of the Dean’s List, and received awards such as the Deutsche Bank Artificial Intelligence prize.

Schedule

The timeframes are only estimates and may vary according to how the class is progressing

Day 1

Choosing the Right Model Architecture

  • Common Architectures: CNNs, RNNs, GANs
  • What’s the right architecture for time series prediction on Bitcoin price data?
  • How do we clean and normalize our time series data?
  • Exercise: Exploring the Bitcoin Dataset and Preparing Data for Model
  • Discussion/Q&A

Break 15 mins

Using Keras as a TensorFlow Interface

  • How does Keras simplify Tensorflow?
  • Using Keras to iterate on deep learning ideas
  • Building our first Keras model and storing it on disk
  • Exercise: Creating a TensorFlow Model Using Keras
  • Discussion/Q&A

Break 15 mins

Training a deep neural network with Keras

  • How do we train a network in Keras
  • Reshaping Time Series Data
  • Predictions and overfitting
  • Exercise: Assembling a deep learning system
  • Discussion/Q&A

Day 2

  • Model Evaluation
  • Loss functions, accuracy, and error rates
  • Introduction to Tensorboard
  • Evaluating our Bitcoin prediction model
  • Exercise: Creating an active training environment
  • Discussion/Q&A

Break 15 mins

  • Hyperparameter Optimization
  • Stacking up more layers in our neural network
  • Exploring different activation functions
  • Using regularization to prevent overfitting
  • Exercise: Optimizing a deep learning model

Break 15 mins

  • Deployment
  • Dealing with online learning and new data points
  • Introducing Docker, Flask, and Redis
  • Deployment and using our app Cryptonic
  • Exercise: Deploying your first deep learning application
  • Discussion/Q&A