README.md

metadata

license: mit
title: sundew_demo
sdk: gradio
emoji: 🚀
colorFrom: indigo
colorTo: indigo
short_description: ' Sundew: Adaptive Energy-Aware Gating Algorithm'
sdk_version: 5.46.1

Sundew Algorithms Interactive Demo

A comprehensive demonstration of the Sundew bio-inspired adaptive gating algorithm with proven 77-94% energy savings across domains.

What This Demo Shows

This interactive demo visualizes how the Sundew algorithm:

1. Multi-Feature Significance Scoring - Combines magnitude, anomaly detection, context, and urgency
2. PI Controller with Hysteresis - Adaptive threshold control with error feedback and stability
3. Energy-Aware Processing - Selective activation achieving substantial energy savings
4. Domain-Optimized Presets - Production-ready configurations for healthcare, IoT, and security domains
5. Statistical Validation - Real-time confidence intervals and performance metrics

Key Features Demonstrated

Production-Ready Algorithm v0.7.1

• Real Performance Data: Uses validated parameters from comprehensive benchmarking
• Multi-Domain Presets: Healthcare, IoT, financial, and security optimizations
• Statistical Rigor: Bootstrap confidence intervals and performance validation
• Energy Efficiency: Proven 77-94% energy savings across 6 datasets

Interactive Visualizations

• Real-time Gating Decisions: Watch the algorithm adapt to different input patterns
• Threshold Evolution: See PI controller maintaining target activation rates
• Energy Savings Tracking: Live monitoring of energy efficiency gains
• Performance Metrics: Precision, recall, F1 scores with confidence intervals

Running Locally

# Install dependencies
pip install -r requirements.txt

# Run the demo
python app.py

Then open your browser to the displayed URL (usually http://localhost:7860).

Deploying to Hugging Face Spaces

1. Create a new Space on Hugging Face
2. Upload these files:
   • app.py
   • requirements.txt
   • README.md
3. Set SDK to "Gradio" and Python version to 3.10+
4. The demo will automatically deploy

Understanding the Visualizations

Main Dashboard: Multi-Domain Performance

• Performance Comparison: Real results across healthcare, IoT, financial domains
• Preset Selector: Try production-ready configurations
• Live Metrics: Energy savings, precision, recall with confidence intervals

Real-Time Processing Panel

• Significance Timeline: Multi-component scoring (magnitude + anomaly + context + urgency)
• Adaptive Threshold: PI controller with error feedback and hysteresis
• Activation Decisions: Green dots show processed events, gaps show energy savings

Statistical Validation

• Bootstrap Confidence Intervals: 95% CIs from 1000 samples (like production validation)
• Performance Trends: Real-time F1, precision, recall tracking
• Domain Benchmarks: Compare against validated production results

Domain-Optimized Presets

Healthcare

• custom_health_hd82: Heart disease optimized (82% energy savings, 0.196 recall)
• custom_breast_probe: Breast cancer with enriched features (77% savings, 0.118 recall)

IoT & Sensors

• auto_tuned: General sensor monitoring (88% savings, 0.500 recall)
• ecg_v1: ECG/cardiac monitoring optimization

Security & Finance

• aggressive: Network security and financial anomaly detection (89-90% savings)
• conservative: Maximum energy efficiency (>92% savings)

Technical Innovation Highlights

Bio-Inspired Adaptive Control

• PI Controller: Maintains stable activation rates despite input variability
• Hysteresis: Prevents oscillation through differential thresholds
• Energy Pressure: Bio-inspired energy management and regeneration

Multi-Component Significance

significance = w_magnitude × magnitude + 
               w_anomaly × anomaly + 
               w_context × context + 
               w_urgency × urgency

Proven Performance

• 77-94% Energy Savings: Validated across 6 real-world datasets
• Statistical Rigor: Bootstrap confidence intervals (95% CI, 1000 samples)
• Production Ready: Hardware integration templates and runtime monitoring

Real-World Applications

Healthcare

• ECG Monitoring: MIT-BIH dataset validation (88.8% energy savings)
• Clinical Decision Support: Heart disease and breast cancer screening

IoT & Edge Computing

• Sensor Networks: Multi-sensor anomaly detection with 88% efficiency
• Edge AI: Energy-constrained processing optimization

Security & Finance

• Network Security: Intrusion detection (89% energy savings)
• Financial Monitoring: Real-time fraud detection and market analysis

Performance Validation

All demo presets are based on comprehensive validation:

• 6 Real Datasets: Healthcare, IoT, ECG, financial, network security
• Statistical Validation: 1000 bootstrap samples with 95% confidence intervals
• Ablation Studies: Component-wise performance analysis
• Adversarial Testing: Robustness against drift, spikes, and noise

Try the demo to see how Sundew achieves production-ready energy efficiency while maintaining critical performance metrics!