A Complete Guide on TensorFlow 2.0 using Keras API

A Complete Guide on TensorFlow 2.0 using Keras API

English | MP4 | AVC 1280×720 | AAC 44KHz 2ch | 133 lectures (13h 5m) | 2.58 GB

Build Amazing Applications of Deep Learning and Artificial Intelligence in TensorFlow 2.0

Welcome to Tensorflow 2.0!

TensorFlow 2.0 has just been released, and it introduced many features that simplify the model development and maintenance processes. From the educational side, it boosts people’s understanding by simplifying many complex concepts. From the industry point of view, models are much easier to understand, maintain, and develop.

Deep Learning is one of the fastest growing areas of Artificial Intelligence. In the past few years, we have proven that Deep Learning models, even the simplest ones, can solve very hard and complex tasks. Now, that the buzz-word period of Deep Learning has, partially, passed, people are releasing its power and potential for their product improvements.

The course is structured in a way to cover all topics from neural network modeling and training to put it in production.

In Part 1 of the course, you will learn about the technology stack that we will use throughout the course (Section 1) and the TensorFlow 2.0 library basics and syntax (Section 2).

In Part 2 of the course, we will dig into the exciting world of deep learning. Through this part of the course, you will implement several types of neural networks (Fully Connected Neural Network (Section 3), Convolutional Neural Network (Section 4), Recurrent Neural Network (Section 5)). At the end of this part, Section 6, you will learn and build their own Transfer Learning application that achieves state of the art (SOTA) results on the Dogs vs. Cats dataset.

After passing the part 2 of the course and ultimately learning how to implement neural networks, in Part 3 of the course, you will learn how to make your own Stock Market trading bot using Reinforcement Learning, specifically Deep-Q Network.

Part 4 is all about TensorFlow Extended (TFX). In this part of the course, you will learn how to work with data and create your own data pipelines for production. In Section 8 we will check if the dataset has any anomalies using the TensorFlow Data Validation library and after learn how to check a dataset for anomalies, in Section 9, we will make our own data preprocessing pipeline using the TensorFlow Transform library.

In Section 10 of the course, you will learn and create your own Fashion API using the Flask Python library and a pre-trained model. Throughout this section, you will get a better picture of how to send a request to a model over the internet. However, at this stage, the architecture around the model is not scalable to millions of request. Enter the Section 11. In this section of the course, you will learn how to improve solution from the previous section by using the TensorFlow Serving library. In a very easy way, you will learn and create your own Image Classification API that can support millions of requests per day!

These days it is becoming more and more popular to have a Deep Learning model inside an Android or iOS application, but neural networks require a lot of power and resources! That’s where the TensorFlow Lite library comes into play. In Section 12 of the course, you will learn how to optimize and convert any neural network to be suitable for a mobile device.

To conclude with the learning process and the Part 5 of the course, in Section 13 you will learn how to distribute the training of any Neural Network to multiple GPUs or even Servers using the TensorFlow 2.0 library.

What you’ll learn

  • How to use Tensorflow 2.0 in Data Science
  • Important differences between Tensorflow 1.x and Tensorflow 2.0
  • How to implement Artificial Neural Networks in Tensorflow 2.0
  • How to implement Convolutional Neural Networks in Tensorflow 2.0
  • How to implement Recurrent Neural Networks in Tensorflow 2.0
  • How to build your own Transfer Learning application in Tensorflow 2.0
  • How to build a stock market trading bot using Reinforcement Learning (Deep-Q Network)
  • How to build Machine Learning Pipeline in Tensorflow 2.0
  • How to conduct Data Validation and Dataset Preprocessing using TensorFlow Data Validation and TensorFlow Transform.
  • Putting a TensorFlow 2.0 model into production
  • How to create a Fashion API with Flask and TensorFlow 2.0
  • How to serve a TensorFlow model with RESTful API
Table of Contents

1 Welcome to the TensorFlow 2.0 course! Discover its structure and the TF toolkit
2 Course Curriculum & Colab Toolkit
3 BONUS 10 advantages of TensorFlow
4 BONUS Learning Path

TensorFlow 2.0 Basics
5 From TensorFlow 1.x to TensorFlow 2.0
6 Constants, Variables, Tensors
7 Operations with Tensors
8 Strings

Artificial Neural Networks
9 Project Setup
10 Data Preprocessing
11 Building the Artificial Neural Network
12 Training the Artificial Neural Network
13 Evaluating the Artificial Neural Network
14 HOMEWORK Artificial Neural Networks
15 HOMEWORK SOLUTION Artificial Neural Networks

Convolutional Neural Networks
16 Project Setup & Data Preprocessing
17 Building the Convolutional Neural Network
18 Training and Evaluating the Convolutional Neural Network
19 HOMEWORK Convolutional Neural Networks
20 HOMEWORK SOLUTION Convolutional Neural Networks

Recurrent Neural Networks
21 Project Setup & Data Preprocessing
22 Building the Recurrent Neural Network
23 Training and Evaluating the Recurrent Neural Network

Transfer Learning and Fine Tuning
24 What is Transfer Learning
25 Project Setup
26 Dataset preprocessing
27 Loading the MobileNet V2 model
28 Freezing the pre-trained model
29 Adding a custom head to the pre-trained model
30 Defining the transfer learning model
31 Compiling the Transfer Learning model
32 Image Data Generators
33 Transfer Learning
34 Evaluating Transfer Learning results
35 Fine Tuning model definition
36 Compiling the Fine Tuning model
37 Fine Tuning
38 Evaluating Fine Tuning results

Deep Reinforcement Learning Theory
39 What is Reinforcement Learning
40 The Bellman Equation
41 Markov Decision Process (MDP)
42 Q-Learning Intuition
43 Temporal Difference
44 Deep Q-Learning Intuition – Step 1
45 Deep Q-Learning Intuition – Step 2
46 Experience Replay
47 Action Selection Policies

Deep Reinforcement Learning for Stock Market trading
48 Project Setup
49 AI Trader – Step 1
50 AI Trader – Step 2
51 AI Trader – Step 3
52 AI Trader – Step 4
53 AI Trader – Step 5
54 Dataset Loader function
55 State creator function
56 Loading the dataset
57 Defining the model
58 Training loop – Step 1
59 Training loop – Step 2

Data Validation with TensorFlow Data Validation (TFDV)
60 Project Setup
61 Loading the pollution dataset
62 Creating dataset Schema
63 Computing test set statistics
64 Anomaly detection with TensorFlow Data Validation
65 Preparing Schema for production
66 Saving the Schema
67 What’s next

Dataset Preprocessing with TensorFlow Transform (TFT)
68 Project Setup
69 Initial dataset preprocessing
70 Dataset metadata
71 Preprocessing function
72 Dataset preprocessing pipeline
73 What’s next

Fashion API with Flask and TensorFlow 2.0
74 Project Setup
75 Importing project dependencies
76 Loading a pre-trained model
77 Defining the Flask application
78 Creating classify function
79 Starting the Flask application
80 Sending API requests over internet to the model

Image Classification API with TensorFlow Serving
81 What is the TensorFlow Serving
82 TensorFlow Serving architecture
83 Project setup
84 Dataset preprocessing
85 Defining, training and evaluating a model
86 Saving the model for production
87 Serving the TensorFlow 2.0 Model
88 Creating a JSON object
89 Sending the first POST request to the model
90 Sending the POST request to a specific model

TensorFlow Lite Prepare a model for a mobile device
91 What is the TensorFlow Lite
92 Project setup
93 Dataset preprocessing
94 Building a model
95 Training, evaluating the model
96 Saving the model
97 TensorFlow Lite Converter
98 Converting the model to a TensorFlow Lite model
99 Saving the converted model
100 What’s next

Distributed Training with TensorFlow 2.0
101 What is the Distributed Training
102 Project Setup
103 Dataset preprocessing
104 Defining a non-distributed model (normal CNN model)
105 Setting up a distributed strategy
106 Defining a distributed model
107 Final evaluation – Speed test normal model vs distributed model

Annex 1 – Artificial Neural Networks Theory
108 Plan of Attack
109 The Neuron
110 The Activation Function
111 How do Neural Networks Work
112 How do Neural Networks Learn
113 Gradient Descent
114 Stochastic Gradient Descent
115 Backpropagation

Annex 2 – Convolutional Neural Networks Theory
116 Plan of Attack
117 What are Convolutional Neural Networks
118 Step 1 – Convolution
119 Step 1 Bis – ReLU Layer
120 Step 2 – Max Pooling
121 Step 3 – Flattening
122 Step 4 – Full Connection
123 Summary
124 Softmax & Cross-Entropy

Annex 3 – Recurrent Neural Networks Theory
125 Plan of Attack
126 What are Recurrent Neural Networks
127 Vanishing Gradient
128 LSTMs
129 LSTM Practical Intuition
130 LSTM Variations

Bonus Lectures