Neural Net Trainer

Build and train neural networks on tabular data directly in your browser using TensorFlow.js.

Features

• Visual model builder — add/remove Dense layers, set units and activation functions
• CSV/JSON data loading — select features and target column
• Classification & Regression — auto-configures loss function and metrics
• Live training chart — watch loss/accuracy update in real-time
• Configurable — epochs, batch size, learning rate, optimizer, validation split
• Export — download trained model as TF.js format
• Test predictions — input feature values, see predictions immediately

How it works

1. Load a CSV/JSON file
2. Select feature columns and target column
3. Choose Classification or Regression
4. Build your model architecture (stack Dense layers)
5. Configure training parameters
6. Click Train — watch the loss chart live
7. Test predictions and export your model

Technical details

• Uses @tensorflow/tfjs for in-browser training
• WebGPU backend when available (10-50x faster than WebGL)
• Automatic one-hot encoding for categorical targets
• Data automatically normalized (min-max) for numeric features