paper-with-me

ANDHRA Module

Ajay N’ Daliparthi Hyper Rectified Activation

2000년 도입 · 논문 1편에서 사용

ANDHRA module splits the neural network into multiple “branches” (representing different paths) that process the same input signal in parallel, each corresponding to different possible outcomes.

출처: ANDHRA Bandersnatch: Training Neural Networks to Predict Parallel Realities

소개 논문: ANDHRA Bandersnatch: Training Neural Networks to Predict Parallel Realities

Activation Functions · General