Discussion on "Identifying and Removing Dead Neurons in Training Neural Networks: Mechanistic Interpretability Part 2"

Abed K · 2024-12-30T19:57:12.148Z

My last blog explained how to correctly initialize the last softmax layer of a neural network to reduce high but incorrect confidence when predicting certain classes. It also covered how to achieve uniform logits, a proper probability distribution, a...

Discussion on "Identifying and Removing Dead Neurons in Training Neural Networks: Mechanistic Interpretability Part 2" | Hashnode

Identifying and Removing Dead Neurons in Training Neural Networks: Mechanistic Interpretability Part 2

Written by

Responses1

Search Hashnode

Identifying and Removing Dead Neurons in Training Neural Networks: Mechanistic Interpretability Part 2

Written by

Responses1