In neural nets, there is often a weight regularization term in the loss function, which ensures that unnecessarily high weights don't occur. For example, if $C(\theta,x)$ is the baseline loss function, then we have the loss $L(\theta,x)=C(\theta,x)+\sum L(\theta_i)$, (where $L$ is the $L_1$ norm or $L_2$ norm usually).

However, there is nothing in principle that prevents us from choosing a more complicated weight regularization term. we can have a generalized weight regularization term $\mathcal L(\theta)$.


  • why not tell our neural network that it should "look like a boolean circuit with 2 inputs in each neuron"? We choose sigmoid activation functions, and use weight regularization that penalizes having more than $2$ non-zero weights for each neuron.

  • why not have a separate neural network look at the original network's weights and its performance, and set weight regularization cost for each weight separately?

Have things in this direction been done? are there benefits? (apart from the obvious computational downsides).

  • 2
    $\begingroup$ The loss function that is the result may be much harder to optimize which can result in getting stuck in local minima and thus decreasing performance. $\endgroup$ – Jan Oct 18 '18 at 14:31
  • $\begingroup$ You use the symbol $L$ in two different ways in your post, which makes it hard to follow. Perhaps you could change the $L$ in the RHS of $L(\theta,x)=\cdots$ into something distinct? $\endgroup$ – Sycorax Dec 18 '18 at 23:59
  • $\begingroup$ I think in elements of statistical learning,they mention regularising to ensure smoothly varying weights ( for images) $\endgroup$ – seanv507 Dec 19 '18 at 0:03

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Browse other questions tagged or ask your own question.