Ayush
Ayush @TensorThrottleX ·
Day 131: Data Science Journey -> Loss Functions: MSE for regression -> Gradient: ∂aL/∂θ_j = (1/n) Σ[(f(θ, x_i) - y_i) * ∂f/∂θ_j] -> Backprop: Chain rule across layers; ∇L via forward act & backward grad ->Learning Rate: Controls step size in GD #ML #DataScience #LearningRate
3
1.4K
dev
dev @zivdotcat ·
Replying to @zivdotcat
The learning rate controls how much to adjust weights with each update. Too high a rate causes overshooting, and too low causes slow convergence. Finding the right balance is crucial for stable training! #LearningRate
1
1
36