Anthony Blaom, PhD comments

Results 815 comments of


                                            Anthony Blaom, PhD

Add new option for exporting learning networks as stand-alone composite model types

@ExpandingMan Thanks for the feedback! > * I wonder if the `machine` with a `Symbol` argument is a necessary pun? I'm not entirely sure what's happening underneath that, and I...

Add new option for exporting learning networks as stand-alone composite model types

@olivierlabayle Did you want to comment on this PR? I realize you might not want to review it as it is quite long, but if you're willing and interested that...

Add new option for exporting learning networks as stand-alone composite model types

Keen to move this along soon. @davnn Have you made any progress testing this PR with OutlierDetection.jl?

Add new option for exporting learning networks as stand-alone composite model types

The MLJ doc update associated with the change proposed here has been posted. The new rewritten Learning Networks section is [here](https://github.com/alan-turing-institute/MLJ.jl/blob/903f0b3a6163a06bec18f1a593ba3b77146e1785/docs/src/learning_networks.md).

Add new option for exporting learning networks as stand-alone composite model types

@davnn Thanks for looking at this. > Where does machine(:detector_name, data...) to machine(getproperty(composite, :detector_name, data...) happen? Whenever you call `fit!(node, composite=composite)` where `node` is a node, then this will interpolate...

Add new option for exporting learning networks as stand-alone composite model types

@olivierlabayle Thanks for taking the time to look over this substantial PR and for your valuable feedback. > Thanks @ablaom for this simplification of the learning network definition process! My...

Add new option for exporting learning networks as stand-alone composite model types

@davnn I'm going to continue the OutlierDetection specific discussion in a new thread which will appear shortly below.

Error during precompilation (ERROR: LoadError: InitError: Evaluation into the closed module `Perceptron` ...)

Replied in issue linked above.

Catch unsupported model types in `BinaryThresholdPredictor`.

@OkonSamuel Can you please review this?

Catch unsupported model types in `BinaryThresholdPredictor`.

@OkonSamuel Thanks for the review. I have replaced the synthetic data with the Pima Indian diabetes data, and replaced f1score with balanced_accuracy. In this case one sees a "statistically significant"...