Deep Learning using Tensorflow Training

Deep Learning using Tensorflow Course:
Opensource since Nov,2015. Roots in Google Brain team.Library for doing Complex Numerical Computation to build machine learning models from scratch.It has scikit-flow similar to scikit-learn for high level machine learning API's. Tensorflow uses Directed Graph as its computational model, similar to Spark. Functions are nodes & edges data. Graph model makes it well suited for deploying Neural Networks.Data flow graph model makes it easily distributed - across CPUs, GPUs & multiple systems.Tensorboard, a visualization software along with Tensorflow makes debugging & analyzing machine learning models really easy.Pre-trained tensorflow model for small devices like mobile, raspberry pi etc makes it highly portable.TensorFlow-Serving is available for deploying pre-trained models in production. Deep Learning is heavily adopted across many companies using TensorFlow.

Deep Learning using Tensorflow Course Curriculum

1. Essential Programming

Introduction to Deep Learning

Introduction to Numpy

Introduction to Tensorflow and Keras

2. Essential basics of Linear Algebra

Solution of Equations, row and column Interpretation

Vector Space Properties

Partial Derivative of Polynomial and Two conditions for Local Minima

Physical Interpretation of gradient (Direction of Maximum Change)

Matrix Vector Multiplication

EVD and interpretation of Eighen Vectors

Linear Independence and Rank of Matrix

Orthonormal Matrices, Projection Matrices, Vandemonde Matrix, Markov Matrix, Symmetric, Block Diagonal

3. Selected topics of Machine Learning

Intuition behind Linear Regression, classification

Grid Search

Gradient Descent

Training Pipeline

Metrics ROC Curve, Precision Recall Curve

Calculating Entropy

4. Basics of Neural Network

Evolution of Perceptrons, Hebbs Principle, Cat Experiment

Single layer NN

Tensorflow Code

Multilayer NN

Back propagation, Dynamic Programming

Mathematical Take on NN

Function Approximator

Link with Linear Regression

Dropout and Activation

Optimizers and Loss Functions

5. Introduction to Convolutional Neural Network

1D and 2D Convolution

Why CNN for Images and speech?

Convolution Layer

Coding Convolution Layer

Learning Sharpening using single convolution Layer in Tensor-Flow

6. Different Layers in CNN pipeline

Convolution

Pooling

Activation

Dropout

Batch Normalization

Object Classification

Creating Batch in Tensorflow and Normalize

Training MNIST and CIFAR datasets

Understanding a pre-trained Inception Architecture

Input Augmentation Techniques for Images

7. Transfer Learning

Finetuning last layers of CNN Model

Selecting appropriate Loss

Adding a new class in the last Layer

Making a model Fully Convolutional for Deployment

Finetune Imagenet for Cats vs Dog Classification.

8. Object Detection and Localization

Different types of problem in Objects

Difficulties in Object Detection and Localization

Fast RCNN

Faster RCNN

YOLO v1-v3

SSD

MobileNet

9. Autoencoders

Image Compression Simple Autoencoder

Denoising Autoencoder

Variational Autoencoder and Reparematrization Trick

Robust Word Embedding using Variational Autoencoder

10. Time Series Modelling

Evolution of Recurrent Structures

LSTM, RNN, GRU, Bi-RNN, Time-Dense

Learning a Sine Wave using RNN in Tensorflow

Creating Autocomplete for Harry Potter in Tensorflow

11. GANs

Generative vs Discrimative Models

Theory of GAN

Simple Distribution Generator in Tensorflow using MCMC (Markov Chain Monte Carlo)

DCGAN,WGANs for Images

InfoGANs, CycleGANs and Progressive GANs

Creating a GAN for generating Manga Art

12. Model Free Approaches in Reinforcement Learning

Model Free Prediction

Monte Carlo Prediction and TD Learning

Model Free Control with REINFORCE and SARSA Learning

Assignment : Implementation of REINFORCE and SARSA Learning in Gridworld

Off policy vs On Policy Learning

Importance Sampling for Off Policy Learning

Q Learning

13. Behavioral Cloning and Deep Q Learning

Understanding Deep Learning as Function Approximator

Theory of Behavioral Cloning and Deep Q Learning

Revisiting Point Collector Example in Unity and

Assignment : Training Cartpole Example via Deep Q Learning

14. Deep Learning in Action

Face Detection using Yolo-v3

Building Autocomplete Feature using RNNs

Real-time Depth Prediction and Pose Estimation

How is Deep Learning used in Autonomous Driver Assistant systems

Tips and Tricks for scaling and easy Deployment of Deep Learning Models

Frequently Asked Questions

What are the modes of training for "Deep Learning using Tensorflow" course?

This "Deep Learning using Tensorflow" course is an instructor-led training (ILT). The trainer travels to your office location and delivers the training within your office premises. If you need training space for the training we can provide a fully-equipped lab with all the required facilities. The online instructor-led training is also available if required. Online training is live and the instructor's screen will be visible and voice will be audible. Participants screen will also be visible and participants can ask queries during the live session.

Will I be provided with any study material during the "Deep Learning using Tensorflow" training?

Participants will be provided "Deep Learning using Tensorflow"-specific study material. Participants will have lifetime access to all the code and resources needed for this "Deep Learning using Tensorflow". Our public GitHub repository and the study material will also be shared with the participants.

What is the pedagogy of zekeLabs?

All the courses from zekeLabs are hands-on courses. The code/document used in the class will be provided to the participants. Cloud-lab and Virtual Machines are provided to every participant during the "Deep Learning using Tensorflow" training.

What is the duration of this course?

The "Deep Learning using Tensorflow" training varies several factors. Including the prior knowledge of the team on the subject, the objective of the team learning from the program, customization in the course is needed among others. Contact us to know more about "Deep Learning using Tensorflow" course duration.

What would be the venue for the "Deep Learning using Tensorflow" training?

The "Deep Learning using Tensorflow" training is organised at the client's premises. We have delivered and continue to deliver "Deep Learning using Tensorflow" training in India, USA, Singapore, Hong Kong, and Indonesia. We also have state-of-art training facilities based on client requirement.

Who is the trainer for "Deep Learning using Tensorflow" training?

Our Subject matter experts (SMEs) have more than ten years of industry experience. This ensures that the learning program is a 360-degree holistic knowledge and learning experience. The course program has been designed in close collaboration with the experts working in esteemed organizations such as Google, Microsoft, Amazon, and similar others.

Can we customize this course based on our requirements?

Yes, absolutely. For every training, we conduct a technical call with our Subject Matter Expert (SME) and the technical lead of the team that undergoes training. The course is tailored based on the current expertise of the participants, objectives of the team undergoing the training program and short term and long term objectives of the organisation.

How can I reach out to you if I have any other queries regarding the "Deep Learning using Tensorflow" course?

Drop a mail to us at [email protected] or call us at +91 8041690175 and we will get back to you at the earliest for your queries on "Deep Learning using Tensorflow" course.