[EyobelG]

Summary

• nn.ts: pluggable optimizer (SGD / Momentum / RMSProp / Adam) with gradient clipping and learning-rate decay, replacing the plain-SGD-only update step. Adds LeakyReLU/ELU/Swish activations, inverted dropout on hidden layers during training, an EarlyStopping helper, and JSON serialize/deserialize for a trained network's weights and biases.
• state.ts / playground.ts / index.html: wires all of the above into the UI — optimizer dropdown, dropout slider, early-stopping checkbox, save/load-model buttons (File/Blob APIs) — plus a dark-mode toggle persisted to localStorage.
• styles.css: dark mode via a page-wide invert filter, with the classification/regression color-coded regions (heatmap, network diagram, dataset thumbnails, loss chart, color legend) counter-inverted so their colors keep their original meaning.

All changes are additive: every new control defaults to the prior behavior (optimizer defaults to sgd, dropout defaults to 0, early stopping defaults off, theme defaults to light), so existing shared URLs/state render identically to before.

Scoped out

• Multi-class classification/cross-entropy+softmax: would require rebuilding the single-output network architecture, the decision-boundary heatmap, and the dataset generators. Left as a follow-up if there's interest.
• Custom CSV/JSON dataset import.

Test plan

• npx browserify src/playground.ts -p "[tsify]" -o dist/bundle.js compiles clean with the project's existing TS 2.9/tsify toolchain.
• Manually exercised in a local browser session: optimizer dropdown (trained the circle dataset with Adam, converged to ~0 loss in <200 epochs), dropout slider, early-stopping checkbox, dark-mode toggle (verified data-encoded colors stay correct), and save-model/load-model round trip (exported JSON, reset the network, re-imported, confirmed identical weights/loss).
• No console errors during the above session.

Happy to split this into smaller PRs per feature if that's easier to review.

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.

[EyobelG]

recheck

[EyobelG]

I signed it!