Skip to content
Datamewby Intura
Start free

Lessons

Introduction to Deep Learning & Neural Networks: all lessons

Neural networks, from a single neuron to the transformer.

6 modules · 21 lessons · 1 free to read now

About this track: outcomes, tools and price

Module 1Neural Network Foundations

4 lessons

What a neural network actually computes, why the nonlinearity is the whole trick, and how the weights learn.

  1. What a Neural Network Actually Computes

    20 min

    A neuron is a weighted sum plus one bending function. Stack a few and a straight line becomes whatever boundary you need.

    Free · read nowRead the lesson Simulation inside

  2. Activation Functions: ReLU, GELU and Friends

    15 min

    Why the sigmoid has all but vanished from hidden layers, and what replaced it.

    Included with the track Simulation inside

  3. Backpropagation Without the Fear

    20 min

    The chain rule, computation graphs, and why one backward pass is enough for millions of weights.

    Included with the track Simulation inside

  4. Your First Training Loop in PyTorch

    25 min

    From tensors to a trained model: forward, loss, backward, step — and how to read the curves.

    Included with the track Simulation inside

Module 2Training Deep Networks

4 lessons

Optimizers, learning-rate schedules, initialisation, normalisation and regularisation — what makes training actually converge.

  1. Optimizers: SGD, Momentum, Adam

    18 min

    What Adam actually does, and when plain SGD still wins.

    Included with the track Simulation inside

  2. Learning-Rate Schedules and Warmup

    15 min

    Cosine decay, warmup, and finding a learning rate in ten minutes.

    Included with the track Simulation inside

  3. Initialisation and Normalisation

    16 min

    Why Xavier and He initialisation exist, and what BatchNorm and LayerNorm fix.

    Included with the track Simulation inside

  4. Regularisation: Dropout, Weight Decay, Early Stopping

    16 min

    Three ways to stop a network memorising, and how to choose between them.

    Included with the track Simulation inside

Module 3PyTorch for Deep Learning

4 lessons

Tensors, autograd, nn.Module, DataLoader and GPU training — enough PyTorch to work on your own.

  1. Tensors and Autograd

    20 min

    Shapes, broadcasting, and how PyTorch records operations to compute gradients.

    Included with the track Simulation inside

  2. Building Models with nn.Module

    18 min

    Composing layers, parameters and train/eval mode without surprises.

    Included with the track Simulation inside

  3. Datasets and DataLoaders

    18 min

    Batching, shuffling and augmentation that do not leave the GPU waiting.

    Included with the track

  4. GPUs and Mixed Precision

    18 min

    Moving training to a GPU, and when float16/bfloat16 is safe.

    Included with the track

Module 4Convolutional Neural Networks

3 lessons

Convolution, pooling and CNN architectures — why weight sharing makes images learnable.

  1. How Convolutions Work

    18 min

    Kernels, stride, padding, and why one filter can be reused across an image.

    Included with the track Simulation inside

  2. CNN Architectures: from LeNet to ResNet

    18 min

    The ideas that lasted — and the skip connection that made very deep networks trainable.

    Included with the track Simulation inside

  3. Transfer Learning

    18 min

    Using a pretrained model when you only have a few hundred images.

    Included with the track Simulation inside

Module 5Sequence Models and Transformers

3 lessons

From RNNs and LSTMs to self-attention — the architecture behind every modern LLM.

  1. RNNs and LSTMs

    16 min

    Processing sequences one step at a time, and why that is hard to train on long sequences.

    Included with the track

  2. The Attention Mechanism

    20 min

    Query, key, value — every token decides which other tokens it needs to look at.

    Included with the track Simulation inside

  3. The Transformer Architecture

    20 min

    Multi-head attention, positional encodings, and the block stacked inside every LLM.

    Included with the track Simulation inside

Module 6Deep Learning in Practice

3 lessons

Debugging training, tracking experiments, and one capstone from raw data to an evaluated model.

  1. Debugging Neural Networks

    18 min

    Overfit one batch first: a recipe for a model that refuses to learn.

    Included with the track Simulation inside

  2. Tracking Experiments

    14 min

    Logging configs, metrics and artefacts so last week's result can be reproduced.

    Included with the track

  3. Capstone: an Image Classifier

    45 min

    From a folder of images to an honestly evaluated model.

    Included with the track