search

airoldilab / ai-sgd

Towards stability and optimality in stochastic gradient descent
5 stars 2 forks source link

Relevant datasets? #3

Closed ptoulis closed 9 years ago

ptoulis commented 9 years ago

I think we should look for some interesting datasets for estimation. Either from the Pascal challenge or from UCI ?

Especially we would like models where explicit would have trouble with; Poisson regression is a prime example.

ptoulis commented 9 years ago

We could also think about Kaggle.

dustinvtran commented 9 years ago

4233ac544b9df20f26a166b99247d208c9a0ba9c runs logistic regression on the raw MNIST dataset. I plan to run some diagnostics in order to experiment with the learning rate, and also compare to the other optimization methods.

ptoulis commented 9 years ago

Excellent! i will give it a try on Friday.