Free course · by Debugshala, Indore
Deep Learning Tutorial
9 free lessons, in the order we teach them in class: start at lesson 1 and work down. Every lesson has an analogy, a diagram, code you can run or copy, a quick quiz and an FAQ.
- 9Lessons
- 2Modules
- ~2 hReading time
- FreeAlways
01Neural Networks
- Neural Network from Scratch in NumPy: Train XOR Step by StepBuild a neural network from scratch in NumPy: neurons, weights, sigmoid, loss, gradient descent and backprop, then train it on XOR and watch the loss fall.
- Activation Functions in Neural Networks: ReLU, Sigmoid, GELUActivation functions explained with NumPy: sigmoid, tanh, ReLU, Leaky ReLU, GELU and softmax, their derivatives, vanishing gradients and which one to pick.
- Backpropagation Explained Step by Step, with NumPy CodeBackpropagation explained step by step: the chain rule on a computational graph, matrix gradients for a tiny network, gradient checking and a toy autograd.
- Gradient Descent and Optimizers: SGD, Momentum and AdamGradient descent optimizers compared in NumPy: batch vs mini-batch SGD, learning rate limits, momentum, RMSProp, Adam and AdamW, plus a PyTorch training loop.
- Batch Normalization and Dropout Explained with NumPy CodeBatch normalization and dropout explained: forward passes in NumPy, running statistics, inverted dropout scaling, train vs eval mode and PyTorch usage.
02Deep Learning Architectures
- Convolutional Neural Networks (CNN) Explained with NumPyConvolutional neural networks explained: a 2D convolution and pooling written in NumPy, stride, padding, output size, channels and a PyTorch CNN for digits.
- RNN and LSTM Explained: Recurrent Neural Networks in NumPyRNN and LSTM explained with NumPy cell steps: hidden state, backpropagation through time, vanishing gradients, LSTM gates, GRU and a PyTorch forecaster.
- Autoencoders Explained: Encoder, Bottleneck and Anomaly UseAutoencoders explained: train a tiny encoder-decoder in NumPy, compare with PCA, detect anomalies with reconstruction error and meet variational autoencoders.
- GANs Explained: Train a Generative Adversarial Network in NumPyGANs explained: generator vs discriminator, the minimax and non-saturating losses, a toy 1D GAN trained in NumPy, mode collapse and a PyTorch training loop.