Deep Learning tutorials and worked examples

Compare controlled updates and diagnose learning behavior.

• Adam vs SGD with momentum on a controlled quadratic

  Compare optimizer updates from the same initial point, record objective traces and distinguish update budgets from wall-clock or general model benchmarks.

  comparison · 4 min read

Trace recurrent states, gates and sequence axes.

• RNN vs LSTM vs GRU: trace states through a short sequence

  Inspect recurrent output shapes, hidden and cell states, parameter counts and order sensitivity using controlled PyTorch modules.

  comparison · 4 min read

Evaluate bottlenecks, reconstructions and their limitations.

• Autoencoders and reconstruction error: compare a linear bottleneck with PCA

  Train a two-to-one-to-two linear autoencoder, compare held-out reconstruction with PCA and expose an unusual point that reconstruction error misses.

  tutorial · 4 min read

Build and debug explicit tensor and training workflows.

• PyTorch train, eval and no_grad: separate behavior from gradients

  Inspect dropout in training and evaluation modes, then verify why eval does not disable gradients and no_grad does not switch the model's mode.

  troubleshooting · 4 min read

• PyTorch vs TensorFlow: which should beginners learn?

  Choose PyTorch or TensorFlow for your first deep learning project. Compare training workflows, Keras 3, tutorial fit and the skills to learn before either.

  comparison · 6 min read

Understand image-model operations and adaptation with worked examples.

• Convolution output shape: work through padding, stride and channels

  Calculate a CNN's spatial dimensions by hand, verify them in PyTorch and repair a mismatched linear layer after flattening.

  tutorial · 4 min read

• Transfer learning vs fine-tuning: the difference and when to use each

  Transfer learning vs fine-tuning explained: how they differ, when to freeze or update pretrained weights, PyTorch code for each, and how LLM fine-tuning fits in.

  comparison · 8 min read

Connect deep learning to the broader modeling toolkit.

• Neural network forward and backward pass on two records

  Trace a two-layer ReLU network, derive one gradient with the chain rule and compare it with PyTorch autograd.

  tutorial · 4 min read

• Machine learning vs deep learning: key differences, with examples

  Compare machine learning and deep learning with practical examples, and see where AI and generative AI fit. Learn when to use each, how much data you need and which skills to study first.

  comparison · 9 min read