Frequency Filter Explorer
See how frequency masking removes noise while preserving signal
📋 Gradient Examples
🎛️ Filter Parameters
💡 How It Works
Apply 2D FFT to convert gradient matrix into frequency components. Each component represents a specific frequency pattern.
Sort all frequency components by their energy (magnitude²). High-frequency noise typically has low energy contribution.
Keep top ρ% of components by energy, zero out the rest. ρ decreases during training (0.95 → 0.50) for progressive denoising.
Apply inverse FFT to get cleaned gradient back in spatial domain. Result: noise removed, essential structure preserved.
🎯 Real-World Impact
Reduces gradient variance by up to 40%, leading to smoother convergence.
Prevents overfitting to high-frequency patterns in training data.
FFT adds ~26% overhead but improves sample efficiency.
📥 Input: Original Gradient
Gradient corrupted with high-frequency noise
Notice: High values (noise spikes) in noisy gradient
📤 Output: Cleaned Gradient
After frequency filtering at ρ = 0.75
Notice: Noise spikes removed, smooth structure preserved
🔬 Frequency Domain Analysis
Full Magnitude Spectrum
After Energy-Based Masking
🧪 Try It Yourself
Experiment 1: Compare Clean vs Noisy
- Select "Clean neural network gradient"
- Note the smooth spectrum (energy concentrated in center)
- Switch to "Gradient corrupted with noise"
- See scattered high-frequency components appear
- Adjust ρ slider - watch noise components get filtered
Experiment 2: Optimal ρ Value
- Select noisy gradient example
- Set ρ = 0.95 (keep almost everything)
- Gradually decrease ρ to 0.50
- Watch noise reduction increase while signal stays intact
- Notice: Too low ρ (<0.50) removes useful structure
Key Insight
Falcon adapts ρ during training: starts at 0.95 (preserve everything early), ends at 0.50 (aggressive denoising late). This progressive filtering balances exploration (early) with exploitation (late).
"In the frequency domain, we see the skeleton of information—
preserving structure while discarding noise, a sculptor's touch."