Error estimate for a universal function approximator of ReLU network with a local connection
This work addresses the need for improved performance in neural networks by providing error estimates for a locally connected architecture, which is incremental as it builds on existing network analysis.
The paper analyzes the approximation error of a ReLU network with local connections, showing that the error depends on depth and width parameters, and suggests this architecture has broader applicability than fully connected networks, including explaining CNNs.
Neural networks have shown high successful performance in a wide range of tasks, but further studies are needed to improve its performance. We analyze the approximation error of the specific neural network architecture with a local connection and higher application than one with the full connection because the local-connected network can be used to explain diverse neural networks such as CNNs. Our error estimate depends on two parameters: one controlling the depth of the hidden layer, and the other, the width of the hidden layers.