Udemy
    •  
    •  
    •  
    •  
    •  
    •  
    •  
    •  
Subscribe
Turn what you know into an opportunity and reach millions around the world.
Learn More
Your cart is empty.
Keep shopping
Deep Learning with TensorFlow and Google Cloud AI: 2-in-1
Rating: 4.2 out of 5(10 ratings)
132 students

Deep Learning with TensorFlow and Google Cloud AI: 2-in-1

Harness the power of deep learning with Google’s TensorFlow!
Created byPackt Publishing
Last updated 9/2018
English

What you'll learn

  • Gain proficiency in building deep learning projects using TensorFlow without any need to delve into writing models from scratch
  • Build a base for TensorFlow by implementing regression
  • Solve prediction and image classification deep learning problems with TensorFlow
  • Tackle the potential of RNN and LSTM neural networks with TensorFlow to solve time series problems
  • Gain hands-on experience designing, training, and deploying your deep learning models with TensorFlow and Keras to handle large volumes of data and complex neural network architectures
  • Design and experiment with complex neural network architectures using low-level TensorFlow while also using TensorFlow’s high level APIs and Keras
  • Scale out training and prediction using different distributed techniques such as data parallelism using GPUs on your local machine and in the cloud using Google Cloud ML Engine

Course content

2 sections • 51 lectures • 8h 8m total length

  • The Course Overview4:26

    This video provides an overview of the entire course.

  • TensorFlow for Building Deep Learning Models4:22

    How to setup TensorFlow on your system and why is there a need to learn Deep Learning using TensorFlow?

    It is important to understand the purpose behind starting with Deep Learning, which is accompanied by TensorFlow installation.

    • Understand why deep learning started being widely used

    • Understand data scientist definition of TensorFlow

    • Install the TensorFlow

  • Basic Syntaxes, Function Optimization, Variables, and Placeholders7:17

    How to build the base for TensorFlow?

    It is usually hard to follow the official TensorFlow documentation so with the description in the video it becomes way easier to learn.

    • Introduce the different aspects of TensorFlow

    • Learn with step by step implementation of basics

    • Build the base for progressing further

  • TensorBoard for Visualization4:12

    Why is there a need to visualize the TensorFlow computation graph and scalers?

    TensorBoard as a visualization tool becomes quite important in case of longer training runs and to understand the computational graph being created using the TensorFlow code, so it is explained in detail here.

    • Start with introducing the need for TensorBoard

    • Start implementation for logging the log summary

    • Trigger the TensorBoard and achieve our target

  • Start by Loading the Imported Dataset3:49

    In this video, we start by importing the Wisconsin state breast cancer dataset

    • Import sklearn and TensorFlow libraries

    • Import the dataset

    • Create three sets – training, validation, and testing

  • Building the Layers of the Neural Network in TensorFlow6:14

    Add the hidden layers to the network, with dropout and fully connected output layer to build the computational graph of your Neural Network.

    • Create the placeholders for to hold data

    • Add layers using fully connected layers functionality of TensorFlow

    • Add global variables initializer for initializing all the variables

  • Optimizing the Softmax Cross Entropy Function3:23

    Using AdamOptimizer we minimize the loss function, which gets us the best mathematical equation for getting the desired output.

    • Create the loss function

    • Create the optimization function

    • Minimize the loss using the optimizer

  • Using DNN Predicting Whether Breast Cancer Cells Are Benign or Not5:08

    We use our network that we built to predict whether the cancer cells are benign or not.

    • Evaluate the loss until the model has trained properly

    • Evaluate the model on validation dataset

    • Test it on the test dataset and get the results

  • Importing the Two Datasets Using TensorFlow and Sklearn API4:23

    In this video we will import the two datasets, first is the MNIST handwritten dataset and the second one is the face dataset.

    • Import NumPy and TensorFlow libraries

    • Import the datasets

    • Create three sets – training, validation and testing

  • Writing the TensorFlow Code to Add Convolutional and Pooling Layers11:21

    We will learn about the convolution to get nice images that our model can learn from the pattern.

    • Understand the CNN architecture

    • Start coding the convolution and pooling layers

    • Connect these layers with the fully connected layer

  • Using tf.train.AdamOptimizer API to Optimize CNN3:43

    In this video, we will calculate loss and optimize the CNN.

    • Use cross entropy function to calculate loss

    • Create the optimization function

    • Minimize the loss using the optimizer

  • Implementing CNN to Create a Face Recognition System3:59

    In this video, we will be creating our face recognition system.

    • Modify the previous computational graph with different variables shapes

    • Define convolution and pooling layers

    • Train the network

  • Understanding the RNN and the Need for LSTM2:24

    Getting to know about how the recurrent Neural Network architecture works.

    • Understand the basics

    • Get to know the architecture

    • Advanced technical aspects

  • Implementing RNN3:11

    Learn to implement RNN using TensorFlow.

    • Import the RNN cell from TensorFlow

    • Learn about BasicRNN and MultiRNN cell

    • Understand what kind of architectures RNN can be built into

  • Monthly Riverflow Prediction of Turtle River in Ontario8:47

    Applying the RNN knowledge on riverflow level dataset.

    • Create the computational graph

    • Train the network

    • Predict the riverflow levels

  • Implement LSTM Project to Predict Decimal Number of Given Binary Representation12:09

    Applying the LSTM knowledge on binary representations.

    • Create the computational graph

    • Train the network

    • Predict the decimal number of the binary representation

  • Encoder and Decoder for Efficient Data Representation3:32

    Here we will learn about the autoencoders

    • View the autoencoder architecture

    • Understand the purpose of the autoencoder

    • Explore the applications of auto encoders

  • TensorFlow Code Using Linear Autoencoder to Perform PCA on a 4D Dataset4:48

    This video focuses on how to construct the network.

    • Load the dataset

    • Create linear autoencoder network

    • Plot the reduced dimensional data

  • Using Stacked Autoencoders for Representation on MNIST Dataset6:28

    Here we will see the steps to construct the network for representation on MNIST dataset using stacked auto encoder.

    • Load the MNIST dataset

    • Stack the autoencoder

    • Plot the input and the encoded images

  • Build a Deep Autoencoder to Reduce Latent Space of LFW Face Dataset5:05

    We will build deep autoencoder with TensorFlow to reduce the latent space of the LFW face dataset.

    • Load the dataset

    • Create the deep autoencoder network

    • Train the autoencoder so as to minimize the loss

  • Generator and Discriminator the Basics of GAN3:16

    Understand the architecture of GANs with properly understanding all the components of it.

    • View the autoencoder architecture

    • Clear your concepts about GANs

    • Explore the applications of auto encoders

  • Downloading and Setting Up the (Microsoft Research Asia) Geolife Project Dataset4:51

    Download the dataset and explore what this dataset is about.

    • Explore the dataset on the Microsoft portal

    • Download the dataset

    • Set it up to use a single person trajectory dataset

  • Coding the Generator and Discriminator Using TensorFlow4:02

    Start coding the generator and discriminators which are core components of GANs.

    • Use variable scope for each of them

    • Create a method each for both

    • Add layers to each of them so that you have two Neural Networks

  • Training GANs to Create Synthetic GPS Based Trajectories10:06

    We will train GANs with TensorFlow to generate synthetic GPS trajectories.

    • Create the computational graph

    • Train the model

    • Visualize the synthetic trajectories with the actual trajectories

  • Hands-on Deep Learning with TensorFlow

Requirements

  • Familiarity with Python programming is required.

Description

Deep learning is the intersection of statistics, artificial intelligence, and data to build accurate models. With deep learning going mainstream, making sense of data and getting accurate results using deep networks is possible. Tensorflow is Google’s popular offering for machine learning and deep learning. It has become a popular choice of tool for performing fast, efficient, and accurate deep learning. TensorFlow is one of the most comprehensive libraries for implementing deep learning.

This comprehensive 2-in-1 course is your step-by-step guide to exploring the possibilities in the field of deep learning, making use of Google's TensorFlow. You will learn about convolutional neural networks, and logistic regression while training models for deep learning to gain key insights into your data with the help of insightful examples that you can relate to and show how these can be exploited in the real world with complex raw data. You will also learn how to scale and deploy your deep learning models on the cloud using tools and frameworks such as asTensorFlow, Keras, and Google Cloud MLE. This learning path presents the implementation of practical, real-world projects, teaching you how to leverage TensorFlow’s capabilities to perform efficient deep learning.

This training program includes 2 complete courses, carefully chosen to give you the most comprehensive training possible.

The first course, Hands-on Deep Learning with TensorFlow, is designed to help you overcome various data science problems by using efficient deep learning models built in TensorFlow. You will begin with a quick introduction to TensorFlow essentials. You will then learn deep neural networks for different problems and explore the applications of convolutional neural networks on two real datasets. You will also learn how autoencoders can be used for efficient data representation. Finally, you will understand some of the important techniques to implement generative adversarial networks.

The second course, Applied Deep Learning with TensorFlow and Google Cloud AI, will help you get the most out of TensorFlow and Keras to accelerate the training of your deep learning models and deploy your model at scale on the Cloud. Tools and frameworks such as TensorFlow, Keras, and Google Cloud MLE are used to showcase the strengths of various approaches, trade-offs, and building blocks for creating, training and evaluating your distributed deep learning models with GPU(s) and deploying your model to the Cloud. You will learn how to design and train your deep learning models and scale them out for larger datasets and complex neural network architectures on multiple GPUs using Google Cloud ML Engine. You will also learn distributed techniques such as how parallelism and distribution work using low-level TensorFlow and high-level TensorFlow APIs and Keras.

By the end of this Learning Path, you will be able to develop, train, and deploy your models using TensorFlow, Keras, and Google Cloud Machine Learning Engine.

Meet Your Expert(s):

We have the best work of the following esteemed author(s) to ensure that your learning journey is smooth:

  • Salil Vishnu Kapur is a Data Science Researcher at the Institute for Big Data Analytics, Dalhousie University. He is extremely passionate about machine learning, deep learning, data mining, and Big Data analytics. Currently working as a Researcher at Deep Vision and prior to that worked as a Senior Analyst at Capgemini for around 3 years with these technologies. Prior to that Salil was an intern at IIT Bombay through the FOSSEE Python TextBook Companion Project and presently with the Department of Fisheries and Transport Canada through Dalhousie University.


  • Christian Fanli Ramsey is an applied data scientist at IDEO. He is currently working at Greenfield Labs a research center between IDEO and Ford that focuses on the future of mobility. His primary focus on understanding complex emotions, stress levels and responses by using deep learning and machine learning to measure and classify psychophysiological signals.


  • Haohan Wang is a deep learning researcher. Her focus is using machine learning to process psychophysiological data to understand people’s emotions and mood states to provide support for people’s well-being. She has a background in statistics and finance and has continued her studies in deep learning and neurobiology.


    Christian and Haohan together they make dyad machina and their focus area is at the interaction of deep learning and psychophysiology, which means they mainly focus on 2 areas:
     - They want to help further intelligent systems to understand emotions and mood states of their users so they can react accordingly
     - They also want to help people understand their emotions, stress responses, mood states and how they vary over time in order to help people become more emotionally aware and resilient

Who this course is for:

  • This learning path is aimed at data science professionals to give them a solid background in how to scale-out deep learning in particular, how to handle large volumes of data and complex neural network architectures and how to deploy their models on the Cloud for production level systems.

Report abuse