[mlpack-git] [mlpack/mlpack] DatasetMapper & Imputer (#694)

[Ryan Curtin]

> +              const T& mappedValue,
> +              const size_t dimension,
> +              const bool transpose = true)
> +  {
> +    // initiate output
> +    output = input;
> +    size_t count = 0;
> +
> +    if (transpose)
> +    {
> +      for (size_t i = 0; i < input.n_cols; ++i)
> +      {
> +         if (input(dimension, i) == mappedValue ||
> +             std::isnan(input(dimension, i)))
> +         {
> +           output.shed_col(i - count);

I think that here is might be faster to collect the list of columns that needs to be dropped, and then use non-contiguous matrix views to extract what you need to keep.  Something like this:

```
std::vector<arma::uword> colsToKeep;
colsToKeep.push_back(0);
colsToKeep.push_back(2);
colsToKeep.push_back(3);

// Only keep columns 0, 2, and 3.
arma::mat output = input.cols(colsToKeep);
```

But before you commit that change, it is probably worth running some quick tests with `mlpack_preprocess_impute` to ensure that speeds things up.  I think you will see a lot of speedup with large matrices.

---
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/mlpack/mlpack/pull/694/files/a8818316a04506530e2269a2e0a32ba2f6a1c83b#r69922436
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.cc.gatech.edu/pipermail/mlpack-git/attachments/20160707/40d7a765/attachment.html>