**UPDATE:**

`mlsauce`

0.4.0 is now available on Pypi:

```
pip install mlsauce
```

and in R:

```
devtools::install_github("thierrymoudiki/mlsauce/R-package")
```

**END UPDATE:**

Last week, I announced a new version of mlsauce (see AdaOpt classification on MNIST handwritten digits (without preprocessing), AdaOpt (a probabilistic classifier based on a mix of multivariable optimization and nearest neighbors) for R, and AdaOpt), with new features. Well… We are still next week!

This **new version contains**:

- A Gradient Boosting algorithm using randomized networks as weak learners
- An implementation of stump decision trees

I’ve been putting work in this new version like… **crazy**. Because it contains some compiled code in C, ported to Python *via* Cython. Hence, so far, version `0.4.0`

is only available through a Git branch called `refactor`

. If you’re interested in a sneak peek, you can install `0.4.0`

by doing:

- Clone the
`refactor`

branch - When cloned, run
`make install`

into the repo - Play with the examples stored in
`/examples`

(tune the hyperparameters)

Check the repo next week, and you’ll see the new changes merged (at least in Python for now, and hopefully for R). More insights on the Gradient Boosting algorithm will be unveiled in a few weeks. I will also talk about the stump decision trees in more details.

**Stay tuned!**