Deep Learning with TensorFlow and PyTorch

Course Overview

Intermediate

Free Course

For programmers who know Python and some basic machine learning and want to understand neural networks from the arithmetic up. You will be able to build, train, debug, tune and save deep learning models in both PyTorch and TensorFlow/Keras, and judge when a neural network is the wrong tool.

Instructor: Jaidev

Sections: 4

Course Content

Section 1: Neural Network Fundamentals

How a neural network turns arithmetic into learned features: perceptrons, backpropagation worked by hand, activation functions, losses and optimisers. 5 lessons, about 60 minutes.

1. What Deep Learning Is and Why It Matters

2. Perceptrons and Multi-Layer Perceptrons

3. Forward and Backward Propagation

4. Activation Functions (ReLU, Sigmoid, Tanh)

5. Loss Functions and Optimizers (SGD, Adam)

Section 2: Building and Training Models

Setting up both frameworks, writing custom models and training loops, and evaluating, tuning and regularising them honestly. 5 lessons, about 60 minutes.

1. Setting Up TensorFlow and PyTorch

2. Custom Neural Networks in TensorFlow and PyTorch

3. Training Loops and Batch Processing

4. Model Evaluation and Hyperparameter Tuning

5. Overfitting and Regularization

Section 3: Advanced Concepts

The layers and habits that keep real training runs stable and reproducible: batch norm and dropout, initialisation and gradient flow, TensorBoard, and saving models that work elsewhere. 4 lessons, about 45 minutes.

1. Dropout, Batch Normalization, and Callbacks

2. Weight Initialization and Gradient Issues

3. Visualizing Training with TensorBoard

4. Saving and Loading Models

Section 4: Mini Project

One complete build on real data: predict California house prices with a neural network in both frameworks, beat or match a strong baseline, and ship an artefact that works in a fresh process. 1 lesson, about 10 minutes.

1. Housing Price Prediction with Deep Learning