Index

A B C D E F G H I L M N R S T U W
All Classes and Interfaces|All Packages

B

backPropagate(double[][]) - Method in interface me.damoebe.architectures.Backpropagation

The method in which the backpropagation process should be managed.

backPropagate(double[][]) - Method in class me.damoebe.architectures.transformer.mha.Head

BACKPROPAGATION IMPLEMENTATION Mathematical Derivation: Forward Pass: 1.

backPropagate(double[][]) - Method in class me.damoebe.architectures.transformer.mha.MultiHeadAttention

Updates All head QKV weights as well as the weights which are used to merge the head matrices

Backpropagation - Interface in me.damoebe.architectures

An interface for backpropagation implemented by each module that supports delta-gradient based weight updates.

biases - Variable in class me.damoebe.architectures.transformer.mha.Head

All biases of the head in one list, (size = inputEmbeddingSize*inputEmbeddingAmount*3) -> used to calculate the queries, keys and values.

Block<H,N> - Class in me.damoebe.architectures.transformer

The Transformer Bock class containing a mha and multiple mlps.

Block(MultiHeadAttention<H>, N) - Constructor for class me.damoebe.architectures.transformer.Block

The main constructor of the Block class

A B C D E F G H I L M N R S T U W
All Classes and Interfaces|All Packages