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.

翻译：我们提出一种新型神经网络计算单元，该单元采用多重偏置结构，挑战了传统感知机架构。该单元强调在信息跨单元传递时保持未受损信息的重要性，通过为每个单元设置专用偏置在后续阶段应用激活函数。通过实证与理论分析，我们证明专注于增加偏置而非权值，有望显著提升神经网络模型的性能。该方法为优化神经网络内部信息流提供了全新视角。完整源代码参见 https://github.com/CuriosAI/dac-dev。