TensorFlow: Artificial Intelligence with TensorFlow: 3-in-1

Name: TensorFlow: Artificial Intelligence with TensorFlow: 3-in-1
Rating: 3.7 (18 reviews)

Begin your journey to build next-generation AI models from scratch with TensorFlow and create your own machine learning

Created byPackt Publishing

Last updated 7/2018

English

What you'll learn

Build custom reusable components for your mobile app and develop native apps for both iOS and Android
Perform animations in your applications using the animation APIs
Test and deploy your application for a production-ready environment
Grasp the concepts of Redux state management to build scalable apps
Add navigation to your App to build UX components for your React Native App
Integrate with Firebase as a data store and learn how to authenticate a user

Course content

3 sections • 61 lectures • 9h 2m total length

The Course Overview7:07
This video provides an overview of the entire course
Machine Learning Basics6:39
Haven’t given any summary of the core machine learning concepts needed to succeed in the course. Solution: providing this overview. We mention what is assumed of the audience and go over some basic concepts as a refresher.
Summarize essential mathematics assumed of the audience
Provide overview of supervised learning and maximum likelihood for training neural networks
Provide overview of feed-forward neural networks and how they are trained
TensorFlow Basics Part 1: Tensors and Variables8:36
Understand Tensor and Variable objects. Solution: explain the concepts of Tensors and Variables in intuitive ways, and explain best practices and common pitfalls.
Show basic syntax and usage of Tensors and Variables
Compare the differences and similarities between Tensors and Variables
Go over best practices and common pitfalls when using Tensors and Variables
TensorFlow Basics Part 2: Graphs and Sessions7:59
TensorFlow Graphs and Sessions are both the two core concepts in TensorFlow, and the two most misunderstood. Solution: clarify their definitions and common misconceptions.
Provide clear definitions for both tf.Graph and tf.Session
Show common usage examples of Graphs and Sessions
Summarize the key ideas of Graphs and Sessions to clarify ambiguities
TensorFlow Basics Part 3: Training, Saving, and Loading8:05
We need to understand the process of training, saving, and loading our models throughout the course. Solution: devote this video to explaining how this process is done.
Build a small neural network from scratch and train it on some toy data
Show how to save our trained model
Show we can reload the saved model and continue training
Convolutional Neural Networks8:44
All neural network computer vision models use convolutional neural networks. In this video we provide a crucial foundational overview required to understand the rest of the section.
Define the task of image recognition that our models need to learn
Provide graphical representation and terminology of the convolution operation
Outline the basic formula defining convolution
Preprocessing, Pooling, and Batch Normalization4:46
There are many modern tricks the audience must know if they want to make competitive models for computer vision. In this video we go over three of the most popular tricks used in computer vision models.
Describe the steps involved in preprocessing images for a neural network
Go over the max pooling operation and TensorFlow implementation
Go over batch normalization and its TensorFlow implementation
Training a CNN on CIFAR-10 – Part 18:17
Now that the audience knows the concepts, they need to see how we implement a full model from scratch. This video shows how to quickly implement TensorFlow models for image recognition.
Provide overview of CIFAR-10 dataset that we will train on
Introduce tf.estimator and tf.data, essential tools for quickly building models
Implement the model’s input pipeline
Training a CNN on CIFAR-10 – Part 211:54
In the previous video, we implemented the model input pipeline. In this video, we implement everything else, combining every s ingle concept in the previous videos to do so.
Show the high-level code that builds the computational training graph
Walk through the code for building the inference portion of the graph
Walk through the training-specific graph nodes needed for our implementation to learn from the data
Embeddings22:14
Every neural language model uses word embeddings. This video ensures the audience understands the related concepts and implementations.
Provide overview of the general concept of word embeddings
Describe GloVe model, a popular model for learning word embeddings
Walkthrough code for preparing and using embeddings in TensorFlow
Recurrent Neural Networks10:34
Nearly all neural language models include RNNs. This video teaches audience about important RNN architectures and implementations.
Introduce components of a basic recurrent neural network
Provide overview of LSTM architecture, which is the most popular RNN in practice
Show how to implement an LSTM-based network in TensorFlow
Bidirectionality and Stacking RNNs4:44
Advanced models will use bidirectional deep LSTM networks for increased performance. This video teaches the audience about these more complex architectures.
Define “unrolling a computational graph”, essential for understanding bidirectional RNNs
Explain bidirectional RNN and deep RNN diagrams
Show preview of TensorFlow implementations for each concept shown
Models for Text Classification – Part 17:57
We’ve only shown the building blocks of language models. This video combines what we’ve learned to walkthrough a full model implementation.
Provide overview of the dataset that we’ll be training our models on
Describe TFRecords data conversion, required for efficient TensorFlow models
Show how the RNN component interacts with the embedding component we built in video 3.1
Models for Text Classification – Part 27:57
We move beyond the simple model of the previous video, with a more advanced implementation that you won’t see in other tutorials. We also provide a sneak peek into TensorBoard usage, as a teaser for the next section.
Walkthrough the advanced model implementation code
Provide insight into subtleties not explained well in TensorFlow documentation
Analyze model accuracy and loss function plots from TensorBoard
TensorBoard5:54
TensorBoard is the most useful tool for visualizing your models. Here we go over how to use it.
Go over each of the main TensorBoard panels
Provide insight into TensorFlow summaries, which generated the TensorBoard contents
Show how to setup and launch TensorBoard
Working with Estimators13:10
We’ve been using Estimators throughout the course, but only the basics. Now we dive deeper into more advanced customizations.
Show how to setup the TensorBoard projector with an Estimator
Show how to implement a custom SessionRunHook for our Estimator
Show how to export our Estimator models
Training Tips12:04
We’ve glossed over the details of training and optimization so far. In this video we go into more detail on the training and optimization process.
Walkthrough custom implementation of optimize loss
Walkthrough more advanced EarlyStoppingHook implementation
Monitor our model performance and gradients with TensorBoard
Debugging Strategies8:05
Debugging TensorFlow programs is daunting at first, but a necessary skill to learn. In this video we introduce the main tools available for debugging and how to use them.
Work through usage of TensorFlow debugger, tfdbg
Show common usage patterns for tf.assert functions
Show common usage patterns for the tf.Print utility
Requirements for ML at Scale9:10
Discuss differences between production and research environments with TensorFlow.
Go over requirements of any production ML system.
Discuss serving and deployment options for TensorFlow in production.
Show examples of distributed TensorFlow and why it’s useful in production.
TensorFlow with C++29:12
Show how to use the C++ API.
Provide instructions for setting up your computer to use the C++ API.
Show basic usage example of C++ API.
Work through example of loading and using saved models in C++.
TensorFlow Serving14:29
Show how to use TensorFlow Serving.
Provide overview of TensorFlow Serving.
Work through example of querying a model server.
Learn more about how gRPC works.
TensorFlow Lite2:58
TensorFlow released a new library for mobile applications, and this video provides an overview.
Describe the components of TensorFlow Lite.
Describe the common workflow for using Tensorflow Lite.
TPUs12:18
Many have heard about Google’s TPUs, but don’t know the details of how they differ from CPU/GPU for deep learning.
Compare differences between CPU and GPU for deep learning
Describe the TPU architecture
Mention a few way of getting access to TPUs
AutoML7:05
Engineers spend a lot of time doing the same types of tasks. This video shows modern techniques for automating some of these tasks.
Provide definitions for AutoML
Go over automated hyperparameter search techniques
Go over automated architecture search techniques
TensorFlow Eager12:10
TensorFlow recently released their new Eager execution mode, which dramatically simplifies how to write TensorFlow code. This video shows how to use it and how it compares with the graph execution mode.
Introduce Eager execution mode and how it differs from graph execution mode
Work through re-implementing video 1.5 in eager execution mode
Course Summary and Next Steps8:10
Summarize what we’ve done throughout the course and share suggestions for continuing with TensorFlow.
Summarize what we’ve done throughout the course
Share problem-solving approaches for TensorFlow projects
Test Your Knowledge

The Course Overview2:26
This video provides an overview of the entire course.
The Current State of Artificial Intelligence2:47
An overview of the current state of deep learning and artificial intelligence.
Discussion on what all algorithms are making a noise
Examples of different learning techniques
What can we expect in the days to come?
Setting Up the Environment for Deep Learning4:17
Setup a virtual environment for deep learning.
Install and setup Anaconda
Create a virtual environment using Conda
Install Tensorflow inside the newly created virtual environment
Deep Learning in Fashion11:19
Build an image classifier to identify Fashion items.
Use Convolutional neural net to classify image
Analyze model performance with TensorBoard
Propose plausible ways to further improve the model
An Intro to Transfer Learning: Skin Cancer Classification7:18
An overview of transfer learning with practical demonstration.
Explore Google’s inception Learning
Reuse learned parameters to solve new classification problems
Train Deep neural networks without GPU
Fundamentals of Object Localization and Detection7:23
The aim of this video is to learn the main concepts behind object localization.
Learn fundamentals of object localization
Explore object localization with Deep Learning techniques
Learn to measure detection accuracy
YOLO(You Only Look Once): Single Shot Object Detection14:39
Learn the state of the art of Object Detection using YOLO.
Look into YOLO 9000
Prepare dataset and annotation for training YOLO
Train an test YOLOv2 using Darkflow
Unravelling Adversarial Learning and Generative Adversarial Nets7:27
An introduction to Generative Adversarial Nets.
Learn about the process of adversarial learning
Learn about Deep Convolutional Generative Adversarial Nets
Look into examples of adversarial learning and DCGANs
Generating Handwritten Digits Using GANs6:03
Replicate the handwritten digits from the MNIST dataset using a GAN.
Download the MNIST dataset
Build and train the discriminator and the generator model
Calculate the discriminator loss for the real image and the generated image and update network weights
Generating New Pokemons Using a DCGAN7:02
Generate new Pokemons using a DCGAN.
Prepare an image dataset of all the different Pokemons
Build a DCGAN following recommendations provided by the DCGAN authors
Train both the generator and the discriminator networks and generate new images
Super-Resolution Generative Adversarial Networks5:45
Reconstruct a low resolution, blurry or pixelated image to increase its resolution.
Download the MNIST dataset and down sample images to prepare the low-resolution versions
Build the SRGAN model
Train the model and generate high resolution images
Setting Up OpenAI Gym1:30
Setup development environment for Reinforcement Learning.
Get to know what is Gym
Install Gym using pip
Introduction to Reinforcement Learning3:07
What is Reinforcement Learning and what are the applications?
Basic understanding of the Reinforcement Learning strategy
Examples of Reinforcement Learning
Simple Q-Learning: Building Our First Video Game Bot6:03
Building a gamebot that solves the FrozenLake game.
Get acquainted with game environments in Gym
Understand the reward mechanism in RL
Understand the Q learning strategy
Deep Q-Learning: Building a Game Bot That Plays the Classic Atari Games5:50
Building a gamebot that learns to play Atari.
Convolution layers for scene understanding
Allow network to train from past experience
Compute Q values during update using a "target" network
Deep Reinforcement Learning with Policy Gradient - AI that Plays Pong4:27
Build an AI that plays pong.
Setup the policy network
Feed raw pixels to the policy neural network
Train the AI
Test Your Knowledge

The Course Overview4:12
This video gives an overview of the entire course.
Installation and Setup8:30
Show to the viewer the computational environment that we are going to use (Anaconda) and how to install TensorFlow-GPU version.
Explain what is the Anaconda Distribution and the problem it solves. Create a new virtual environment
Explain the necessary software for installing the GPU version of TensorFlow
Install TensorFlow-GPU and test the installation
Defining Layers for Image Recognition12:54
Present image classification as the main problem in Computer Vision, introduce Convolutional Neural Networks and its main components: Convolutional Layers and Pooling Layers.
Introduce Computer Vision and the Image Classification problem
Explain what a Convolutional Neural Network is explain its components
Provide a visual understanding of Convolutional and Max-Pooling layers
Building an Image Classifier with CNNs11:13
Show how to use TensorFlow to build a CNN model and use it for an image classification task.
Introduce the MNIST Fashion dataset
Show and explain the code used for building an image classifier
Evaluate the CNN classifier and show the results
Building Better CNNs with Regularization5:10
Show how to build better CNN models using regularization and different architectures of layers.
Show a different option of architectures of layers that can be used in CNN
Show how to add dropout regularization to a CNN
Compare the results of the new model in term of classification accuracy
Transfer Learning13:42
Explain what Transfer Learning is and show an example on how to use this technique to solve a Computer Vision problem.
Explain what Transfer Learning is
Introduce the image classification problem
Explain and show the code to implement Transfer Learning in Keras
The Intuition Behind RNNs11:06
Explain the main ideas related to RNNs and why they are the main deep learning models to learn from sequences.
Explain what is a sequence and the different cases of models with sequences
Explain at high level what is a RNN
Present the code for building a toy RNN
Time Series Forecasting with RNN10:10
Provide an example of building a RNN using the BasicRNNcell to predict a time series.
Present the BasicRNNcell class
Present the problem and explain the dataset
Explain the code and implementation of the RNN for predicting a time series
Producing Word Embeddings for NLP Tasks11:57
Vectorization of text is one of the most common tasks in Natural Language Processing. In this video we explain the word embedding technique and the gensim library for implementing it.
Explain some of the vectorization techniques for NLP
Explain the intuition and goal of word embeddings
Present the gensim library and the Word2Vec class for producing embeddings
Processing Text Sequences with LSTM Networks10:33
Introduce the LSTM cell which is a more powerful alternative to BasicLSTMCell, explain the idea behind LSTM cells and use them to produce a review classifier.
Explain the main characteristics of LSTM cells
Explain the code for producing the word embeddings
Present the implementation of the LSTM network and the results of the classifier
Guessing Correlations from Scatter Plots12:57
Talk about some of the achievements of artificial intelligence systems and present a narrow task in which we can build an AI program that can beat humans: guessing correlations from scatter plots.
Present some of the areas where AI perform better than humans
Explain the problem we want to solve: guessing the correlation coefficient from scatter plots
Present the code for building the AI program to solve the proposed problem
Introduction to Generative Adversarial Networks8:54
Talk about some of the achievements of artificial intelligence systems and present a narrow task in which we can build an AI program that can beat humans: guessing correlations from scatter plots.
Present some of the areas where AI perform better than humans
Explain the problem we want to solve – guessing the correlation coefficient from scatter plots
Present the code for building the AI program to solve the proposed problem
Creating Images with GANs15:02
Present one implementation of GANs and explain the code for building a GAN that can produce new images of shoes.
Present the code for building and training a GAN
Explain how to obtain new images with the trained model
Give some advice on training GANs
Sequence to Sequence Models6:05
Introduce the idea of sequence to sequence models and explain at a high level how to implement one of such models for language translation.
Introduce sequence to sequence models
Explain the encoder – decoder setup
Explain the different processes of training the model and getting inferences
Building a Language Translator13:49
Present one implementation of a sequence to sequence model that will translate short sentences from English to Spanish using character-level model.
Present the dataset and the problem to solve
Explain the implementation of the model for training
Explain the implementation of the model for inference
Key Concepts in Reinforcement Learning7:59
Introduce the field of Reinforcement Learning and how it differs from the traditional ML and some of the key concepts used this sub-field of AI.
Give a brief introduction to the field of Reinforcement Learning
Explain the concepts of Agent, Actions, Environment, State, Reward and their relationship
Introduce some other concepts like Policy and Episodic and Continuous tasks
A Simple Environment and Basic Policies8:29
Introduce OpenAI and one concrete example of all the concepts discussed in the previews video.
Introduce and talk about the library OpenAI
Show the code for creating an environment
Show how to apply some basic policies
Training a Neural Network Policy8:42
Present and explain one methodology for training a neural network for obtaining a policy to solve the CartPole environment.
Present the Neural Network to be trained
Explain the methodology of adjusting the gradients based on rewards
Explain the training loop for the neural network
Using an Intelligent Agent3:41
Show how to use the agent trained in the last video and how it performs in the CartPole environment.
Present the code for loading the trained network
Demonstrate how the “intelligent agent” behaves
Give a summary of the section
Test You Knowledge

Requirements

Knowledge of Data Science

Description

Google’s TensorFlow framework is the current leading software for implementing and experimenting with the algorithms that power AI and machine learning. Google deploys TensorFlow for many of its products, such as Translate and Maps. TensorFlow is one of the most used frameworks for Deep Learning and AI. This course will be your guide to understand and learn the concepts of Artificial intelligence by applying them in a real-world project with TensorFlow.

This comprehensive 3-in-1 course is a practical approach to deep learning and deep reinforcement learning for building real-world applications using TensorFlow. Learn how models are made in production settings, and how to best structure your TensorFlow programs. Build models to solve problems in Computer vision, Natural Language Processing, Reinforcement Learning, Finance, and more!

Contents and Overview

This training program includes 3 complete courses, carefully chosen to give you the most comprehensive training possible.

The first course, Learn Artificial Intelligence with TensorFlow, covers creating your own machine learning solutions. You’ll embark on this journey by quickly wrapping up some important fundamental concepts, followed by a focus on TensorFlow to complete tasks in computer vision and natural language processing. You will be introduced to some important tips and tricks necessary for enhancing the efficiency of our models. We will highlight how TensorFlow is used in an advanced environment and brush through some of the unique concepts at the cutting edge of practical AI.

The second course, Hands-on Artificial Intelligence with TensorFlow, covers a practical approach to deep learning and deep reinforcement learning for building real-world applications using TensorFlow. This course will take you through all the relevant AI domains, tools, and algorithms required to build optimal solutions and will show you how to implement them hands-on. You’ll then be taken through techniques such as reinforcement learning, heuristic searches, neural networks, Computer Vision, OpenAI Gym, and more in different stages of your application. You’ll learn how TensorFlow can be used to analyze a variety of data sets and will learn to optimize various AI algorithms. By the end of the course, you will have learned to build intelligent apps by leveraging the full potential of Artificial Intelligence with TensorFlow..

The third course, TensorFlow 1.x Deep Learning Recipes for Artificial Intelligence Applications, covers recipes for Computer vision, Natural Language Processing, Reinforcement Learning, Finance, and more! Build models to solve problems in different domains such as Computer vision, Natural Language Processing, Reinforcement Learning, Finance, and more. Taking a Cookbook approach, this course presents you with easy-to-follow recipes to show the use of advanced Deep Learning techniques and their implementation in TensorFlow. After taking this tutorial you will be able to start building advanced Deep Learning models with TensorFlow for applications with a wide range of fields.

By the end of the course, you’ll begin your journey to build next-generation AI models from scratch with TensorFlow and create your own machine learning solutions.

About the Authors

Brandon McKinzie is an NLP engineer/researcher and lover of all things associated with machine learning, with a particular interest in deep learning for natural language processing. The author is extremely passionate about contributing to research and learning in general, and in his free time he’s either working through textbooks, personal projects, or browsing blogs related to ML/AI.
Saikat Basak is currently working as a machine learning engineer at Kepler Lab, the research & development wing of SapientRazorfish, India. His work at Kepler involves problem-solving using machine learning, researching and building deep learning models. Saikat is extremely passionate about Artificial intelligence becoming a reality and hopes to be one of the architects of the future of AI.
Alvaro Fuentes is a Data Scientist with an M.S. in Quantitative Economics and a M.S. in Applied Mathematics with more than 10 years' experience in analytical roles. He worked in the Central Bank of Guatemala as an Economic Analyst, building models for economic and financial data. He founded Quant Company to provide consulting and training services in Data Science topics and has been a consultant for many projects in fields such as: Business, Education, Psychology and Mass Media. He also has taught many (online and on-site) courses to students from around the World in topics such as Data Science, Mathematics, Statistics, R programming, and Python. Alvaro Fuentes is a big Python fan; he has been working with Python for about 4 years and uses it routinely to analyze data and make predictions. He also has used it in a couple of software projects. He is also a big R fan, and doesn't like the controversy between what is the “best” R or Python; he uses them both. He is also very interested in the Spark approach to big data, and likes the way it simplifies complicated topics. He is not a software engineer or a developer but is generally interested in web technologies. He also has technical skills in R programming, Spark, SQL (PostgreSQL), MS Excel, machine learning, statistical analysis, econometrics, and mathematical modeling. Predictive Analytics is a topic in which he has both professional and teaching experience. He has solved practical problems in his consulting practice using Python tools for predictive analytics and the topics of predictive analytics are part of a more general course on Data Science with Python that he teaches online.

Who this course is for:

Data science enthusiast looking to achieve the power of Artificial Intelligence for developing machine learning solutions using TensorFlow, then this course is what you need.
Developers and aspiring Data Science professionals who would like to develop their AI techniques to create smart and robust applications.
Data Analysts, Data Scientists, Data Engineers, Software Engineers, and anyone working with Python and data who wants to perform Machine Learning on a regular basis and use TensorFlow to build Deep Learning models.

TensorFlow: Artificial Intelligence with TensorFlow: 3-in-1

What you'll learn

Explore related topics

Course content

Learn Artificial Intelligence with TensorFlow26 lectures • 4hr 20min

Hands-on Artificial Intelligence with TensorFlow16 lectures • 1hr 37min

TensorFlow 1.x Deep Learning Recipes for Artificial Intelligence Applications19 lectures • 3hr 5min

Requirements

Description

Who this course is for: