Build an interactive neural-net explainer

Build a single self-contained HTML page that teaches a beginner how a small neural network learns. It must be a genuinely interactive explainer, not a static diagram. Requirements: a visual diagram of a tiny feed-forward network (input → hidden → output) with neurons and weighted connections; a "Train" button that runs gradient descent on a simple task (e.g. learning XOR or fitting a curve) and animates the network learning; a live loss curve that updates as training progresses and visibly trends down; a control to change the learning rate and a "Reset" button; and clear, friendly labels so a first-time learner understands what they're seeing. Design matters most here — the spacing, alignment, color, and overall polish should look like a professional educational product, not an AI-generated page. Make it beautiful and genuinely useful for someone learning machine learning 101. All CSS in a `<style>` tag and all JS in a `<script>` tag — no external libraries, images, or fonts. Return ONLY a single complete self-contained HTML file, no markdown fences, no explanation.