Data Augmentation

Implement deterministic data augmentation transforms for 1D signals.

Given a 1D signal, apply the following transforms based on the transform parameter:

1. “flip”: Reverse the signal: [1,2,3] -> [3,2,1]
2. “scale”: Multiply by a factor: signal * factor
3. “shift”: Circular shift by amount positions to the right. Positive shifts to the right: [1,2,3,4], shift=1 -> [4,1,2,3]
4. “noise”: Add a fixed noise pattern. Given a noise tensor, return signal + noise.

Input:

• signal: 1D tensor of shape (L,)
• transform: string, one of “flip”, “scale”, “shift”, “noise”
• params: dict with transform-specific parameters:
  • flip: {} (no params)
  • scale: {“factor”: float}
  • shift: {“amount”: int}
  • noise: {“noise”: tensor of shape (L,)}

Output: Augmented signal of shape (L,).