GUI Tools

Interactive applications for SOEN model design and analysis

GUI Tools

SOEN Toolkit provides several interactive GUI applications for model creation, parameter visualization, and result analysis. These tools make it easy to work with complex superconducting neural networks without deep knowledge of the underlying configuration formats.

Available GUI Applications

All GUI tools are launched using Python's module system:

Model Creation GUI

Interactive interface for designing SOEN model architectures.

Launch Command

python -m soen_sim_v2.model_creation_gui

Features

  • Drag-and-drop layer creation - Add SingleDendrite, DoubleDendrite, RNN, LSTM, and other layers
  • Visual connection editor - Draw connections between layers with automatic constraint checking
  • Parameter configuration - Set layer dimensions, solver types, and physics parameters
  • Architecture validation - Real-time validation of model topology and parameter compatibility
  • YAML export - Generate complete configuration files for training
  • Template library - Start from common architectures (feed-forward, recurrent, hybrid)

Layer Types Available

Physical SOEN Layers:

  • SingleDendriteLayer
  • DoubleDendrite1Layer
  • DoubleDendrite2Layer
  • ScalingLayer

Virtual Layers:

  • InputLayer
  • RNNLayer, LSTMLayer, GRULayer
  • MinGRULayer
  • LeakyRNNLayer
  • ClassifierLayer

Connection Types

  • Dense connections - Fully connected weight matrices
  • Sparse connections - Configurable sparsity patterns
  • Custom connectivity - Upload custom connection masks
  • Internal connections - Recurrent dynamics within layers

Use Cases

  • Rapid prototyping - Quickly test new architectures
  • Educational tool - Understand SOEN model structure visually
  • Collaboration - Share visual model designs with colleagues
  • Architecture search - Systematically explore model variations

Physical Parameter Mapping GUI

Interactive tool for visualizing and configuring superconducting device parameters.

Launch Command

python -m soen_sim_v2.physical_mappings_gui --port 5001

Features

Parameter Visualization:

  • 3D phase space plots - Visualize φ, g, s parameter relationships
  • Time evolution curves - See how parameters change during simulation
  • Power consumption maps - Understand energy usage patterns
  • Noise visualization - View effect of stochastic perturbations

Interactive Configuration:

  • Parameter sliders - Adjust γ₊, γ₋, dt, and other values in real-time
  • Source function selector - Choose and configure physics-based activations
  • Noise configurator - Set up stochastic and deterministic perturbations
  • Constraint editor - Define parameter bounds and relationships

Physical Units:

  • Unit conversion - Switch between physical and dimensionless quantities
  • Physical constants - Configure Φ₀, Ic, ωc for your specific devices
  • Device modeling - Map abstract parameters to real superconducting circuits

Applications

  • Device design - Optimize parameters for specific superconducting devices
  • Physics education - Understand superconducting dynamics visually
  • Parameter tuning - Find optimal settings for your applications
  • Validation - Verify that parameters are physically realistic

TensorBoard Results Viewer

Enhanced interface for analyzing SOEN training results and model behavior.

Launch Command

python -m soen_sim_v2.view_tb_results_gui

Features

Training Analysis:

  • Loss curves - Training and validation loss over epochs
  • Metric tracking - Accuracy, perplexity, bits per character
  • Learning rate schedules - Visualize LR changes during training
  • Gradient analysis - Histogram of gradient magnitudes

SOEN-Specific Visualizations:

  • Layer state evolution - How layer states change over time
  • Power consumption - Energy usage per layer and total
  • Physical parameter drift - Track learnable physics parameters
  • Connection weight heatmaps - Visualize learned connectivity patterns

Model Behavior:

  • State space trajectories - Phase plots of layer dynamics
  • Activation patterns - Which neurons are active when
  • Time series analysis - Temporal behavior of recurrent layers
  • Comparison tools - Compare multiple training runs

Advanced Features:

  • Interactive plots - Zoom, pan, select time ranges
  • Export capabilities - Save plots and data for publications
  • Custom metrics - Add your own analysis functions
  • Real-time monitoring - Watch training progress live

Screenshots

Model Creation Interface

SOEN Toolkit GUI Tools Screenshots

Complete GUI suite: Model Creation, Physical Parameter Mapping, and TensorBoard Results Viewer

🎯 All GUI Tools Integrated

The SOEN Toolkit GUI suite provides comprehensive tools for model design, parameter visualization, and results analysis - all accessible through simple Python module commands.

Installation & Setup

Prerequisites

Follow the instructions found at the GitHub repository: https://github.com/greatsky-ai/soen-toolkit.

Next Steps

View Examples → API Reference → Physical Modeling →