TL;DR We propose to locally decorrelate the feature weights of CNNs. When
the proposed method, which we call
OrthoReg, is used to regularize the 40
layers of Wide Residual Networks, we obtain state of the art results on CIFAR,
and SVHN. Here is an example of the effects of our local regularizer:
Average nearest neighbor distance
I encourage the reader to continue reading for better understanding of this
Using a script to run multiple tasks in background can be very handy but it can also be anoying when we press crtl+c to interrupt and the background processes keep working. There are many approaches to solve this but I have found the cleanest one is to simply trap the...
After plugging the second monitor, I could not find the correct resolution in nvidia-settings and it was impossible to do it manually. After having to solve this problem for the second time, I post it here so to remember.
Look at nvidia-settings for the name of the monitors (In...
I decided to share this script I was using in order to accelerate the training of neural nets making them to gather data in parallel. So I wrote some tests to check it performs well its basic functions and commented the code a little bit.
Convert Matconvnet models to Caffe with matconv2caffe.
There is already a script to import caffe models to matconvnet thanks to @vedaldi. However, we needed the oposite so I made a script that converts the .caffemodel and extracts a deplotment .prototxt model definition and the average image. Currently it...