Tamper with hyperparameters as a simple neural network trains.
Just go to the live demo.
- backpropagation from scratch in nn.js, self-contained but unsophisticated
- fragile bit-packing to run inference as a fragment shader
- jQuery to smooth over the usual papercuts
- Andrej Karpathy's ConvNetJS demo graphs where the input grid lands in each layer's output space.
- Tensorflow's playground graphs each neuron's output as a function of the 2D input space, including parts where the sample data doesn't cover.