A neural network laboratory running directly in your browser.
Build a deterministic network and prepare the complete experiment step by step.
Local file
Project import preview
- Format
- —
- Version
- —
- Experiment
- —
- Topology
- —
- Seed
- —
- Inputs / outputs
- —
- Patterns / active
- —
- —
- Trained network
- —
- Loss history
- —
- Errors
- 0
- Warnings
- 0
The file may contain the complete dataset and trained weights. Check it for sensitive data before sharing.
Step 01
Network definition
Core diagnostics
Direct normalized input to URAX Core. This tool does not use the ranges of defined variables.
Step 02
Experiment definition
Name the experiment and describe the meaning of inputs and outputs in their original units.
Input layer
Input definitions
Output layer
Output definitions
Step 03
Training Data
Enter inputs and expected outputs in their original units. Normalization happens only when training is prepared.
Local processing · max. 2 MB / 10,000 rows
CSV import preview
Column mapping
Map every required target exactly onceImport check
No problemsFirst rows preview
Values are not normalizedThe dataset changes only after confirmation.
There are no training patterns yet
Add the first row. Default values use the minima from the experiment definition.
Step 04
Training
The network learns locally in a Web Worker. One epoch processes every active valid pattern exactly once.
Run parameters
Training settings
Live status
Metrics
- Waiting for the first run.
Step 05
Results
Evaluation uses the current network state and displays predictions in the original user units.
Original units
Summary metrics
MSE, RMSE, and MAE include all active valid patterns and all outputs.
Normalized space
Loss history
The chart will be available after the first training step.
Active valid patterns
Dataset results
Recalculate results to display the table.
Current network snapshot
New prediction
Enter values and calculate a prediction.