RowProcessor issue

[Mohammed-Ryiad-Eiadeh]

After construction 'var Fprocessor = new HashMap<String, FieldProcessor>();
Fprocessor.put("feature.", new DoubleFieldProcessor("feature."));
Fprocessor.put("algo.", new DoubleFieldProcessor("algo."));'
it gives my a warning like ' WARNING: Row 949 empty of features, omitting'.
but when i use the exact field name (the name of the feature), it works fine.
i have checked the regular expression .*, but i don't know what is the problem

Is your question about a specific ML algorithm or approach?
Please describe the ML algorithm, with appropriate links or references.

Is your question about a specific Tribuo class?
List the classes involved.

System details

• Tribuo version : Tribuo-4.2.1
• Java version (if appropriate) : JDK 17 (corretto)
• OS/Architecture (if appropriate) : windows 11

Additional context
Add any other context or screenshots about the question

[Craigacp]

If you want to use a regex to generate the field processors at loading time then you're supplying the Mapping argument to the wrong constructor. You need to use this one:

    public RowProcessor(List<FieldExtractor<?>> metadataExtractors, FieldExtractor<Float> weightExtractor,
        ResponseProcessor<T> responseProcessor, Map<String,FieldProcessor> fieldProcessorMap,
        Map<String,FieldProcessor> regexMappingProcessors, Set<FeatureProcessor> featureProcessors)

Then pass in Mapping to regexMappingProcessors and supplying Collections.emptyMap() to fieldProcessorMap. RowProcessor really needs to use a builder to construct it as it's rather complicated, but we haven't got around to designing auto-generated builder support for Configurable classes, and didn't want to add one-off builders to classes. We should probably revisit that decision in the case of RowProcessor though as it's really very confusing.

[Mohammed-Ryiad-Eiadeh]

well, now everything is work as it should be.
i hope you will keep evolve Tribuo library and deliver such strong models like FMtrainer.
i think of building K-Medoid Clustering algorithm and ask you if it is suit for Tribuo

[Craigacp]

I think K-Medoids would be a useful algorithm to add into Tribuo, though we don't have plans to add it ourselves at the moment. Contributions to add it are welcome. The sparse by default nature of Tribuo's datasets will make it slightly trickier, but it should look fairly similar to the K-Means implementation, and can probably use the same KMeansModel class at prediction time with a different trainer class.

[Mohammed-Ryiad-Eiadeh]

sounds well, so i should understand K-Means model in Tribuo first.
sounds very well.