Linear Layer Internals
Input X (with Bias Trick)
×
Model Weights W (m and c)
=
Output (Forward Pass)
True Weights W (m and c)
Computation Trace:
1. $X \cdot W$ (Matmul) ✅
2. Current Epoch: 0.00
2. MSE Loss: 0.00
3. Gradient: m:0, c:0
4. $W = W - \eta \cdot \nabla L$ ✅
1. $X \cdot W$ (Matmul) ✅
2. Current Epoch: 0.00
2. MSE Loss: 0.00
3. Gradient: m:0, c:0
4. $W = W - \eta \cdot \nabla L$ ✅