Corrado Mio

It is possible to compile the library using MSYS2 GNU C++. However there is a problem on how GNU C++ and Microsoft CL handle the template instantiation. The problem is...

JavaSymbolSolver unable to resolve an inner class defined in a base class

5

Given the following classes: ``` public class ConvolutionLayer { public static class Builder {} } ``` ``` public class Convolution2D extends ConvolutionLayer { } ``` ``` public class Test {...

[ENH] Better integration of NeuralForecaster

11

In sktime there are no models based on NN. I found 3 libraries: 1) Darts 2) Python Forecasting 3) NeuralForecaster 4) Flow Forcast (2022) 5) tsai (2024) (if there are...

[BUG] Timeseries models as reduction of regression models are "confusing" and with "unreasonable" dependences between "fit" and "predict"

3

This is not exactly a BUG, but an "confused" implementation problem. -- I think the architecture related to TS models based on REGRESSION models is a little confusing: too variants...

[BUG] `predict` returns wrong index if inner mtypes are `numpy` - can we add custom output handling?

6

I don't know IF this is a bug OR a possible extension. We suppose to have 1. Xtrain, ytrain, Xpred, dataframes with a PeriodIndex/DatetimeIndex 2. a timeseries class, extending 'sktime.forecasting.base.BaseForecaster',...

To use a dedicates namespace for the logging.

1

To use the Python 'logging' package to write logs is good. BUT it is BETTER (and simple) to use a dedicated namespace, for example "castle.algorithm." for the algorithms, and "castle.general"...

[BUG] fh_in_fit breaks the standard library interface

3

**Describe the bug** You suppose to implement the code to compare the behaviour of ALL time series models available. The DEFAULT code is: ``` model = Model(init params) modelf.fit(y, X)...