sgd

sgd copied to clipboard

dustinvtran

Assumption using Fisher information to give variances, but we should also be able to check whether or not SGD is in the asymptotic region

dustinvtran

How to handle feature selection, like `glmnet`? Should allow for something such as elastic net to do L1/L2 regularization. - [x] GLMs (explicit) - [x] GLMs (implicit) - [ ]...

dustinvtran

add namespace sgd, and sub-namespace things

2

I like this approach to avoid names clashing during implementation. For instance (arbitrarily chosen), see Stan's approach as in the file https://github.com/stan-dev/stan/blob/develop/src/stan/mcmc/sample.hpp.

dustinvtran

Flags for constructing design matrix is ignored

4

- [ ] `subset`: a subset of data points; can be a parameter in sgd.control - [ ] `na.action`: how to deal when data has NA; can be a parameter...

lantian2012

Allow user to specify generic loss function to do SGD on

2

This cannot be efficiently implemented in Rcpp. Rcpp cannot interface with the R function in any way other than calling it in R each time. Hence for now, we are...

dustinvtran

Since we're doing scalable computation, we should work with scalable I/O, storage, etc. packages. That is, we'd like to be able to run SGD on data sets with memory larger...

dustinvtran

One-dim-eigen broken

1

It's awful. ``` R library(sgd) # Dimensions N

dustinvtran

Unit testing

5

Add unit tests to `tests/` folder. Follow similar structure as https://github.com/hadley/dplyr/tree/master/tests. See http://r-pkgs.had.co.nz/tests.html

dustinvtran

dustinvtran