Anthony Blaom, PhD

icon indicating a website link ablaom.github.io icon indicating an email [email protected]

icon company University of Auckland icon location Auckland icon location Mathematician and Data Scientist, PhD Caltech • Expertise: Differential Geometry, Dynamical Systems, Fluid Mechanics, Statistics, Machine Learning, Julia

Results 815 comments of Anthony Blaom, PhD

Control caching of composite models

> Of course I guess there is the possibility of adding a hyperparameter to the composite model that can be transferred after to the machine definitions. This would probably solve...

Improve querying for measures

The problem with this, of course, is that measures should ideally be conceptualised as independent of MLJ altogether, and machines are an MLJ concept.

`serializable` does not remove data implicitly stored in partial functions, leading to scaling with data size and potential privacy breaches

I've created a test PR referenced above of [for-a-0-point-20-release](https://github.com/JuliaAI/MLJBase.jl/tree/for-a-0-point-20-release) onto `dev` which hopefully will reproduce the fail.

`serializable` does not remove data implicitly stored in partial functions, leading to scaling with data size and potential privacy breaches

Perhaps the resampling update to `Stack` is responsible.

`serializable` does not remove data implicitly stored in partial functions, leading to scaling with data size and potential privacy breaches

Ah-haaa! Yes that makes sense. @olivierlabayle Thanks for taking time to diagnose this problem. This is an interesting case I had never thought of: data being stored in partial functions....

Issue informative warning if user tries to use `evaluate!(::MLJModel, ...)` instead of `evaluate(::MLJModel, ...)`.

Didn't know about this feature, thanks. That's a pattern I've wondered about. Guess it has to wrapped in a block to test the feature actually exists, as it is still...

Implement efficient data utils methods for Tables.MatrixTable

Agreed. Waiting for: - [ ] resolution of https://github.com/JuliaData/Tables.jl/issues/281

Request: Separate Out Resampling Strategies Interface

I think it's a good idea and reasonable request. Another possibility would be to use https://github.com/JuliaML/MLDataPattern.jl instead.

Failure to precompile in Julia 1.7.1

I am not able to reproduce but did have a different issue that was resolved with ```julia import Conda Conda.update() ``` You may have to add `Conda` to your environment....

Add this kaggle attempt to external resources

> it relies on a dataset that we can't serve. Oh well. A link would be good.