Concepts
Start with the intuition, inspect a figure, then work through the math. Each lesson links back to places where the idea is used.
Learning and optimization
- Multilayer perceptron — weighted sums, nonlinear responses and intermediate representations.
- Backpropagation — calculate sensitivity by following dependencies backward.
- Gradient descent — turn a gradient into a parameter update.
Signals and neurons
- EEG and frequency bands — voltage differences, time scales and interpretation.
- Spike train — events, rates and timing.
- Leaky integrate-and-fire neuron — accumulation, leak, threshold and reset.
Evaluating predictions
- Confusion matrix — see which classes get confused.
- Cohen’s kappa — agreement relative to a declared marginal-frequency baseline.