Igor Babuschkin
Igor Babuschkin
I noticed that when solving a `GradientProblem`, the cost reported is only the second-to-last one. For example, when minimizing something like `f = (a + b)^2` and starting at `a...
We want to make everware more usable, especially for users that don't know how docker and jupyter work. Here are some key points from our meeting with Ricardo: - [...
Through #49 we already have some very basic support for saving your progress from within everware. It currently allows you to - fork the pulled repo if you don't have...
The performance of the network on GPUs seems to be lagging behind the CPU performance. I suspect that this is because the 2D convolution isn't designed to work efficiently if...
We've discussed the fact that one-hot encoding the input to the network is kind of weird, and that it would be more natural to use the waveform as a single-channel...
I've started to play around with the [MagnaTagATune dataset](http://mirg.city.ac.uk/codeapps/the-magnatagatune-dataset). There's a small change that needs to be made to the code when training on this dataset: Because it uses mp3...
When only a certain range of the data is used for the fit, or when a range has been cut out, the PDFs need to be adjusted accordingly by removing...
We should be able to easily sample data from the model.
Currently we're using [OptimizeResult](http://docs.scipy.org/doc/scipy/reference/generated/scipy.optimize.OptimizeResult.html#scipy.optimize.OptimizeResult) from scipy when returning results. The idea is to have an object that basically behaves like a dictionary, but also pretty-prints the results and various useful...
We should be able to create models where the data variables have different lengths. We should then be able to append them and put their combination into a model. This...