Increasing biases can be more efficient than increasing weights
This offers an alternative perspective on neural network optimization, potentially improving efficiency for AI practitioners, though it appears incremental in scope.
The authors tackled the problem of optimizing information flow in neural networks by introducing a novel computational unit with multiple biases, showing that increasing biases rather than weights can significantly enhance model performance.
We introduce a novel computational unit for neural networks that features multiple biases, challenging the traditional perceptron structure. This unit emphasizes the importance of preserving uncorrupted information as it is passed from one unit to the next, applying activation functions later in the process with specialized biases for each unit. Through both empirical and theoretical analyses, we show that by focusing on increasing biases rather than weights, there is potential for significant enhancement in a neural network model's performance. This approach offers an alternative perspective on optimizing information flow within neural networks. See source code at https://github.com/CuriosAI/dac-dev.