Raw ECG Data: signal array

→ ECGDataset: to PyTorch Tensors, unsqueeze(1)

→ DataLoader: batch=128, shuffle

Section 2: Dilated TCN Architecture (The Model)

Layered 1D Convolutions (d=1,2; k=7,5)

ReLU + BatchNorm

Adaptive Avg Pooling (128 units)

Linear Layer → 5-class logits

Forward Pass: Cross-Entropy Loss

Backward Pass: loss.backward()

Optimizer: Adam (optimizer.step())

Logging: Epochs & Loss tracking