Tony Bagnall

Results 139 comments of Tony Bagnall

looking at the tags, I think the only ones we dont want to take from the series transformer are _tags = { "input_data_type": "Collection", "output_data_type": "Collection", "X_inner_type": "numpy3D", # or...

also, we enforce calling fit by setting is fitted to true in the base class fit not the constructor

so I think we have some issues with threading about cloning/not cloning and the old problem of where to set n_jobs. I propose that we get it in without threading...

I'll push some changes, but they are open for discussion

there is in fact a problem calling _fit, since the X_inner_type of collections and series are fundamentally different, will unfortunately have to call base class fit

> Actually, with forecasting more officially deprecated now, I can probably remove the legacy check function. Can keep for now if people disagree. I vote remove it

we can do these things before going to 1.0.0. Personally I think 1.0.0 should come at the end of the grant

I'll leave this issue open but I dont think we have enouigh information to call it a bug

I'm closing this as we do not let have forecasters with prediction intervals, or indeed any forecasters (see #2266). plot_series needs redesigning to take numpy arrays in any case