TMats / survey

Summary of Paper Survey
https://tmats.github.io/survey/
16 stars 2 forks source link

Equilibrium Propagation: Bridging the Gap Between Energy-Based Models and Backpropagation #71

Open TMats opened 6 years ago

TMats commented 6 years ago

https://arxiv.org/abs/1602.05179

TMats commented 6 years ago

実装 https://github.com/bscellier/Towards-a-Biologically-Plausible-Backprop Yoshua Bengio - Deep learning and Backprop in the Brain (CCN 2017)(youtube) https://www.youtube.com/watch?v=W86H4DpFnLY

TMats commented 6 years ago

Abustruct

1. Introduction

2. Continuous Hopfield Model再訪(もっと生物学的妥当性のある誤差逆伝播としてのEquilibrium Propagation)

3. Energy basedなモデルの学習の枠組み

4. 関連研究

5. Implementation & Results

6. Discussion

TMats commented 6 years ago

次に読む

Early Inference in Energy-Based Models Approximates Back-Propagation

STDP as Presynaptic Activity Times Rate of Change of Postsynaptic Activity Approximates Backpropagation