2 votes
Accepted

What values are actually used for the kernel when using a Conv2D layer?

The values are learned during the training. Initially, they are given random values, then by means of back-propagation and gradient descent, the values of the kernels are adjusted during several ...
  • 19k
2 votes

Which desktop hardware is best for DL?

This question is complex because GPUs' power and memory are not linear. I had some projects where some additional power or memory would have a much better result from one GPU to another, without any ...
1 vote
Accepted

How to bias a neural network towards one category in binary classification?

Adjusting your classification threshold is probably the easiest solution to try first. By default, you're likely classifying probabilities above 0.5 as class 1. Lower threshold would mean more entries ...
  • 656
1 vote

Which desktop hardware is best for DL?

I built a PC specifically for ML last year and from my research at that time here are some of my conclusions. My datasets are quite a bit smaller than yours so I didn't spend too much time thinking ...
  • 31

Only top scored, non community-wiki answers of a minimum length are eligible