Skip to content

Latest commit

 

History

History
155 lines (111 loc) · 3.95 KB

File metadata and controls

155 lines (111 loc) · 3.95 KB

Neural Network Quick Start Guide

How to Use the Neural Prediction Feature

1. Open the App

open apps/ai-tools/recursive-self-portrait.html

or visit: http://localhost:8000/apps/ai-tools/recursive-self-portrait.html

2. Start Observation

Click "Begin Observation" button

3. Move Your Cursor

Move your cursor around for at least 10-15 movements to collect training data

4. Enable Neural Mode

Check the "Use Neural Prediction" checkbox

5. Watch It Learn!

  • Neural Confidence meter will start at 0% and increase as you move
  • Network visualization will show neurons activating
  • After ~100-200 movements, confidence should reach 60-80%

Quick Test Pattern

Move your cursor in a circle:

  1. Start at the center
  2. Move clockwise in a circle
  3. Complete 2-3 full circles
  4. Watch the neural confidence rise to 70%+
  5. The ghost cursors should start predicting your circular motion!

Toggle Comparison

Try this to compare neural vs heuristic:

  1. Move in circles with neural mode ON
  2. Note the prediction accuracy
  3. Toggle neural mode OFF
  4. Keep moving in circles
  5. Compare which method predicts better!

What to Look For

Good Signs

  • ✓ Neural confidence increasing (0% → 70%+)
  • ✓ Prediction trails following your cursor
  • ✓ Network visualization neurons lighting up
  • ✓ Smooth 60 FPS performance

If Something's Wrong

  • Neural confidence stuck at 0%? → Make sure you moved at least 11 times
  • Checkbox not working? → Check browser console for errors
  • Canvas blank? → Scroll down in sidebar to see it
  • Lagging? → Try reducing recursion depth

Key Features

Feature Location What It Shows
Neural Confidence Stats Panel 0-100% training quality
Prediction Mode Stats Panel "Neural" or "Heuristic"
Neural Toggle Below Depth Slider On/Off checkbox
Network Viz Below Fingerprint Live network activations
Prediction Trails Main Viewport Colored prediction paths

Pro Tips

  1. Smooth movements → Higher confidence (70-90%)
  2. Erratic movements → Lower confidence (30-50%)
  3. Patterns → Neural learns them quickly
  4. Random motion → Heuristic may work better
  5. Toggle during movement → See real-time comparison

Keyboard Shortcuts

  • Space - Start/Stop observation
  • R - Reset session
  • W - Toggle webcam
  • (Tab to checkbox) - Toggle neural mode

Troubleshooting

Problem Solution
Confidence not increasing Move cursor more (need 11+ actions)
Can't find neural toggle Scroll down sidebar, it's below depth slider
Canvas shows "Awaiting data" Move cursor with observation started
Performance issues Lower recursion depth to 3-5
Toggle not working Refresh page, try again

Performance Tips

For best performance:

  • Keep recursion depth at 5-7
  • Don't enable webcam and neural simultaneously (can, but more CPU)
  • Close other browser tabs
  • Use Chrome or Firefox for best performance

Understanding the Visualization

Network Canvas

  • Bright neurons = High activity
  • Dim neurons = Low activity
  • Purple neurons = Output (when neural mode active)
  • Lines = Connections between layers

Confidence Meter

  • <30% = Still learning
  • 30-70% = Getting better
  • 70%+ = Good predictions
  • 90%+ = Excellent (on regular patterns)

What's Next?

Try these experiments:

  1. Move in different shapes (circles, squares, zigzags)
  2. Change speed (slow vs fast)
  3. Mix patterns (circle then straight line)
  4. Compare neural vs heuristic on each
  5. See which patterns neural learns fastest

Documentation

For more details:

  • IMPLEMENTATION_SUMMARY.md - Feature overview
  • NEURAL_NETWORK_IMPLEMENTATION.md - Technical details
  • NEURAL_TESTING_GUIDE.md - Comprehensive testing

Enjoy exploring how AI learns your movement patterns in real-time!