Dilum Aluthge

Detect usage of non-public names?

2

Would it be possible to detect uses of non-public names? E.g. `using Foo: bar` where `Foo.bar` is not public. ## Example If I have: ```julia module MyPackage export f public...

Fix `predict_particles`, and implement the Brier score

15

Fixes #93 Fixes #117 Related to #52 Please note that this PR makes multiple breaking changes.

Hyperparameters should be a mutable struct (not a named tuple)

6

That way we can do hyperparameter optimization with MLJ.

Multinomial logistic regression: `predict_particles` throws `MethodError`

5

`predict_particles` throws a `MethodError` when I try to use it in the multinomial logistic regression example. The error is: `ERROR: MethodError: no method matching &(::Particles{Bool,1000}, ::Particles{Bool,1000})` Full output: ```julia julia>...

Implement a loss function for classification models

7

We currently have an example of a loss function for regression models. Specifically, we implement the root mean squared error. However, we don't currently have an example of a loss...

We need to fix the implementation of `MLJModelInterface.predict` for classifiers

5

In #91, I added an incorrect implementation of `MMI.predict` for classifiers. This allowed me to finish the pipeline, do cross validation, add additional tests, etc. But we should fix the...

Reorganization of the predictor types

6

Currently, we only have a single predictor type: ```julia struct SossMLJPredictor{M}

Shorter printing of Soss models in stack traces

5

Whenever the Soss or SossMLJ model types show up in stack traces, it is quite verbose, and makes the stack trace hard to read. Here's an example. Look e.g. at...

Bayesian neural networks

1

I think that Bayesian neural networks would be a natural successor to our tutorials on Bayesian GLMs. As far as the neural network library to use, I think [Flux](https://github.com/FluxML/Flux.jl) makes...