Berent Å. S. Lunde

**Resolves** #56 **Approach** Create an enum for the type of loss function, in this way avoiding the string-type attribute and hopefully resolving the issue. **Pre review checklist** - [ ]...

Efficient serialization of agtboost model objects

1

#### Resolves #57

The human-readable text-serialization is one of the most expensive ways to serialize. See the comment in https://github.com/Blunde1/agtboost/issues/55#issuecomment-1187996654. Take a look at more efficient and robust ways of serialization.

The `agtboost` model class contains the type of loss-function as an `std::string` type attribute. This is problematic (for memory and garbage collection reasons) when working with a model object from...

Release 0.9.3

1

Resolves #50

``` The CRAN policy contains - Packages should not attempt to disable compiler diagnostics, nor to remove other diagnostic information such as symbols in shared objects. yet packages MatchIt agtboost...

Use `testthat` As a minimum, write tests for - `gbt.save()` and `gbt.load()` - `gbt.train()` - `predict()`

Both `xgboost` and `lightgbm` implements monotone constraints. This is important to practitioners. An introduction to a possible implementation is [given here](https://towardsdatascience.com/how-does-the-popular-xgboost-and-lightgbm-algorithms-enforce-monotonic-constraint-cf8fce797acb)

There are no problems for the information criterion in regards to implementing a fast histogram-based algorithm. This should make the sorting which is `nlog(n)` be linear in `n`.