Paragon AI Live Demo

Experience type-generic neural networks with WebGPU acceleration and distributed growth

Loading Paragon WASM Engine...

WebGPU Acceleration

GPU vs CPU benchmarking

Type-Generic

Float32, Int32, Uint32 networks

Network Growth

Self-evolving architectures

ADHD Evaluation

Advanced performance metrics

1 Performance Benchmarking

Test CPU vs GPU performance across different network types and activation functions

Benchmark Settings

Single Forward Pass Timing

Each test runs one forward pass for precise GPU vs CPU measurement (like the working index.html demo)

Performance Results

Run a benchmark to see CPU vs GPU performance comparisons

About the Benchmark:

  • Tests single forward pass performance on different activation functions
  • Compares CPU vs WebGPU acceleration performance
  • Evaluates type-generic network efficiency (Float32, Int32, Uint32)
  • Results show milliseconds per forward pass and speedup ratios

2 Network Growth Experiment

Create a neural network and watch it evolve through distributed micro-network optimization

Network Configuration

Define the initial network structure

Training Data

Growth Parameters

Waiting for Paragon WASM to load...

3 Network Visualization & Results

View your network architecture and track performance metrics

Current Network Architecture

Create a network to see its structure

-

Total Layers

-

Parameters

0

Growth Events

-

ADHD Score

How It Works

1. Micro-Network Surgery: Extract trainable sub-networks from checkpoint layers

2. Distributed Training: Parallel optimization across multiple worker threads

3. ADHD Evaluation: Performance scoring using Accuracy Deviation Heatmap Distribution

4. Network Evolution: Reintegrate improved architectures back into the main network