Multi-Layer Perceptron for Age Regression
Date:
Built a custom MLP from scratch using NumPy only, and applied L-BFGS-B optimization.
(GitHub Link)
Applied L-BFGS-B optimization, grid search, and 5-fold cross-validation for tuning.
Reduced loss from 1681.80 to 91.31 by implementing He/Xavier initialization, gradient clipping, and custom backpropagation.