Arthur Pignet
Arthur Pignet
The [PR283](https://github.com/SubstraFoundation/distributed-learning-contributivity/pull/283) set a new optimizer, which shown better results. However, I didn't test it with Sequential, nor stratified data.
Yes it is. The split of the data between partners is label agnostic, but it is not he case of the shuffling/corruption Basically, the only type of labels accepted is...
Indeed PVRL is a contributivity method where an agent trained by reinforcement learning (policy gradient) would choose a subset of partners at every epoch. Currently one mpl object is created,...
I think that the MAX_BATCH_SIZE should be related to the GPU memory available, so it will be dataset dependant
I deleted PVRL, Moving-functions, Add-Imdb-dataset[...] and dvrl. All these branches had been either dropped or rebased in an other branch, which has been merged
Great to see that the target had been found ! @natct10 did you have the time to remove the commit ? Can we close this issue ?
> By the way, we really should separate code from its outputs (reports), which could be hosted on this open science oriented platform https://osf.io/. Besides, this would be totally relevant...