DNN · Deep Neural Networks
B.Tech Final Year · Elective Course

Deep Neural Networks

Twenty-four lectures that build the mathematics of deep learning from first principles — from a single perceptron to backpropagation, CNNs, LSTMs, self-attention and transformers, energy-based models, and autoencoders. Every lecture pairs formal derivations with a small hand-worked numerical example, an interactive visualization, and runnable code.

24 lectures Worked numerical examples Interactive visualizations Downloadable Python code
▶ Start with Lecture 1 💾 Browse all code downloads

Module A Foundations of Neural Networks

Lecture 01

Course Introduction & The Deep Learning Landscape

What deep learning is, why it works now, and the roadmap for this course.

 Lecture 02

Machine Learning Foundations Recap

Supervised/unsupervised learning, train/test splits, overfitting, bias-variance.

 Lecture 03

The Perceptron & Biological Inspiration

From biological neurons to the McCulloch-Pitts model and the linear perceptron.

 Lecture 04

Feedforward Neural Networks & Forward Propagation

Stacking layers, matrix formulation of the forward pass.

 Lecture 05

Activation Functions

Sigmoid, tanh, ReLU, Leaky ReLU — math, gradients, and failure modes.

 Lecture 06

Probability, Loss Functions & Cross-Entropy

Random variables, MLE, KL divergence, and where cross-entropy comes from.

 Lecture 07

Gradient Descent & Stochastic Gradient Descent

Optimizing weights: batch GD vs. SGD, learning rate, convergence.

 Lecture 08

Backpropagation: The Core Algorithm

The chain rule, applied systematically — full derivation and hand-worked example.

 Lecture 09

Advanced Optimizers

Momentum, AdaGrad, RMSProp, and Adam.

Module B Evaluation & Practical Implementation

Lecture 10

Evaluation Metrics

MSE/MAE, confusion matrix, precision/recall/F1, ROC & PR curves.

 Lecture 11

Building Neural Networks in Python

From-scratch NumPy network, then the same model in Keras.

Module C Convolutional Neural Networks

Lecture 12

CNNs: Motivation & Architecture

Why dense networks fail on images, and the convolutional alternative.

 Lecture 13

Convolution, Padding, Stride & Pooling

The arithmetic of feature maps, with a hand-computed convolution.

 Lecture 14

CNN in Practice

Flattening, fully-connected layers, softmax, and a full Keras CNN.

Module D Sequence Models

Lecture 15

Recurrent Neural Networks

Modeling sequences: hidden state recurrence and unrolling in time.

 Lecture 16

Vanishing/Exploding Gradients

Why long sequences break vanilla RNN training.

 Lecture 17

LSTM Networks: Gates & Memory

Forget, input and output gates — a full numeric gate walkthrough.

 Lecture 18

LSTM Variants

BiLSTM, stacked BiLSTM, encoder-decoder architectures, and GRU.

 Lecture 19

Word Embeddings & Tokenization

One-hot, Word2Vec (CBOW/Skip-gram), BPE and WordPiece.

Module E Attention & Transformers

Lecture 20

Attention & Self-Attention (Transformers)

Scaled dot-product attention, multi-head attention, and BERT — worked step by step.

Module F Unsupervised & Energy-Based Models

Lecture 21

Self-Organizing Maps & K-Means

Competitive learning, topology-preserving maps, and centroid clustering.

 Lecture 22

Boltzmann Machines, EBMs & DBNs

Energy functions, RBMs, contrastive divergence, and deep belief networks.

Module G Autoencoders

Lecture 23

Autoencoders: Architecture & Training

Encoder-decoder-bottleneck, reconstruction loss, sparse & denoising variants.

 Lecture 24

Advanced Autoencoders & Applications

Contractive, stacked, deep & convolutional autoencoders; anomaly detection.

Resources Code & Downloads

All lectures

💾 Code & Resources Library

Download every lecture's Python demo, plus Colab/Kaggle links for the heavier notebooks.