Page 12 - Demo

Many different regression methods could be used to reconstruct the synthetic source
• Polynomial regression:
• Simple and well-known
• Requires high-order (n > 10) polynomials to fit complex signals
• Prone to divergence near edge of signal, making calibration of edge channels difficult
• K-nearest neighbor algorithm:
• For each point 𝑥", average k nearest points together
• Low complexity – only the single parameter k
• Reconstructed source tends to be jagged and “stepped”
• Averaging many different subsets together mostly smooths out curve, but creates high uncertainty in reconstruction
Presenter: Gabriel Player
12