This page of personal notes was started because I had to look up ‘loss function’.
Comparisons: PyTorch vs TensorFlow (2024 May 8)
model.summary()
: the None
values in the
Output Shapes of the layers are a reminder that the model expects
the input to have a batch size as the outermost dimension
(ref)
Loss function | Error | Used as a cost function in minimization |
ReLU (rectified linear unit) | Rectifier | faster than sigmoid and tanh (Nair & Hinton, 2010; Krizhevsky et al., 2012) and also fixes the ‘vanishing gradient problem’ (https://www.baeldung.com/cs/sigmoid-vs-tanh-functions) |