On Approximation Capabilities of ReLU Activation and Softmax Output Layer in Neural Networks

Open in new window