Welcome to Neural Network from Scratch in TensorFlow! In this 2-hours long project-based course, you will learn how to implement a Neural Network model in TensorFlow using its core functionality (i.e. without the help of a high level API like Keras). You will also implement the gradient descent algorithm with the help of TensorFlow's automatic differentiation. While it’s easier to get started with TensorFlow with the Keras API, it’s still worth understanding how a slightly lower level implementation might work in tensorflow, and this project will give you a great starting point for the same.

Course Objectives In this course, we are going to focus on two learning objectives:

How to implement a neural network model from scratch using TensorFlow.

Solving a multi-class classification problem using the neural network implementation.

By the end of this course, you will be able to implement a Neural Network model in TensorFlow using its core functionality. You will also implement the gradient descent algorithm with the help of TensorFlow's automatic differentiation. Finally, you will be able to use this implementation to solve a multi-class classification problem.

Course Structure This course is divided into 3 parts:

Course Overview: This introductory reading material.

Neural Network from Scratch in TensorFlow: This is the hands on project that we will work on in Rhyme.

Graded Quiz: This is the final assignment that you need to pass in order to finish the course successfully.

Project Structure The hands on project on Neural Network from Scratch in TensorFlow is divided into following tasks:

Task 1: Introduction Introduction to the project.

Introduction to the Rhyme interface.

Importing relevant libraries and helper functions.

Task 2: Initializing Neural Network Creating the Neural Network class.

Understanding the init function.

Setting up initial parameter values.

Task 3: Forward Pass Creating a forward pass function.

Task 4: Computing Loss and Updating Parameters Using the cross entropy loss with logits.

Updating parameters for all the layers.

Task 5: Predict and Info Functions Understanding the pre-written info function.

Creating a predict function with the help of the forward pass.

Task 6: Training on Batch Creating the main training mechanism.

Implementing gradient descent with automatic differentiation.

Task 7: Training on Complete Set Breaking down data-set in batches.

Breaking down the training process in epochs and steps.

Task 8: Application Applying the neural network model to solve a multi-class classification problem.

Creating model instance and setting up hyperparameters.

Training the model.

Task 9: Results Plotting the training results.

A look at predictions on the test set.