Closed zhao1157 closed 4 years ago
In calculating loss, why do we need an additional L2_loss? Is it needed for all models, including the models included in benchmark_cnn?
L2 loss helps avoid overfitting. It's a form of weight regularization, and it useful for most models. Here's a blog post that describes it.
In calculating loss, why do we need an additional L2_loss? Is it needed for all models, including the models included in benchmark_cnn?