[IGNITE-11664] [ML] Use Double.NaN as default values for missing values in Vector

• Type: Improvement
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Component/s: ml
• Labels: stability
• Ignite Flags: Docs Required
• Assignee: Alexey Zinoviev
• Reporter: Alexey Zinoviev
• Created: 01/Apr/19 11:49
• Updated: 29/Dec/20 08:50

Description

Currently, we use 0.0 value for default values in vectors if a value is missing. But this way contradicts to preprocessors politics where for missing values Double.NaN is using. Moreover, Double.NaN is a more convenient value for missing feature values.