Can I hold 2 batches of dlnetwork gradients and update network parameters in 1 operation?

Neeta Dsouza answered . 2025-07-30 20:16:57

function state = aggregateState(state,factor)

    numrows = size(state,1);
    
    for j = 1:numrows
        isBatchNormalizationState = state.Parameter(j) =="TrainedMean"...
            && state.Parameter(j+1) =="TrainedVariance"...
            && state.Layer(j) == state.Layer(j+1);
        
        if isBatchNormalizationState
            meanVal = state.Value{j};
            varVal = state.Value{j+1};
            
            % Calculate combined mean
            combinedMean = gplus(factor*meanVal);
                   
            % Caclulate combined variance terms to sum
            combinedVarTerm = factor.*(varVal + (meanVal - combinedMean).^2);        
            
            % Update state
            state.Value(j) = {combinedMean};
            state.Value(j+1) = {gplus(combinedVarTerm)};
           
        end
    end
end

Not satisfied with the answer ?? ASK NOW