MNIST, Snake, and Word Vectors.
Open demos →Meeting 1How a neural network moves informationNeurons, weights, biases, ReLU, and forward propagation.
Open lesson →Meeting 2How a model measures errorTargets, signed errors, squared loss, and total loss.
Open lesson →Meeting 3How do we reduce loss?Try training a model by hand and discover why choosing the right changes gets difficult.
Open lesson →Meeting 4Backpropagation, step by stepTrace sensitivities backward, then update a network across four examples.
Open lesson →Meeting 5From one update to trainingExplore learning rates, initialization, and repeated training.
Open lesson →Meeting 6Learning from digitsTrain a small MNIST network, explore batches and epochs, and evaluate unseen digits.
Open lesson →Meeting 7Is it actually learning?Compare training and validation, explore model layers, and inspect mistakes.
Open lesson →