Use SignalWeave Modern with a real sample project

1. Open SignalWeave Modern

You can use SignalWeave Modern in two ways:

• Open the browser version at /app/.
• Download the desktop release for Windows, Linux, or macOS from the homepage downloads section.

When the app opens, the left side is the main workspace, the right side is the control stack, and the bottom shows the console and error graph.

2. Load the sample project

Use the sample project file:

seventeen-patterns-7x7-modern.swproj.json

1. Click 📂 Load.
2. Select the sample project file.
3. Confirm that the network graph updates and the Patterns tab shows the embedded dataset.

The project file already contains the network definition, patterns, weights, training sessions, and workspace state.

3. Train the sample

Open the Training tab and review the main settings:

• Learning rate: how aggressively weights change
• Momentum: how much previous update direction is preserved
• Error threshold: target error for stopping logic
• Learning steps: size of one visible training block

Click Train #1. The app creates a training session checkpoint, updates the error graph, and appends a structured report to the console.

If you click Train #2 later, training continues from the current weights rather than restarting from scratch.

4. Test patterns and inspect the graph

1. Open the Tests tab.
2. Click Test all to evaluate the full pattern set.
3. Select a pattern from the list.

The selected pattern updates the main network graph and the inspector shows inputs, targets, outputs, delta, hidden activations, and current error.

Use Test selected when you want to refresh only one pattern instead of the full set.

5. Inspect weights

Open the Weights workspace tab on the left.

Choose the source checkpoint, the layer you want to inspect, and the view mode: Heatmap, Magnitude, or Values.

This is the fastest way to compare what changed between checkpoints and understand which connections are strong, weak, positive, or negative.

6. Run analysis

Open the Analysis workspace tab and choose a mode:

• Output clustering
• Hidden-state clustering
• Compatibility summary
• Time series
• Surface plot

Then choose the source checkpoint and click Run. You can copy the result, save text or CSV output, or save chart modes as PNG.

7. Save your work

Use 💾 Save to update the current project file or 🖫 Save As to create a new project file.

The Modern project format keeps all important state together, including network settings, patterns, weights, completed cycles, and training sessions.

Next steps