Pawel Flajszer Slip-box Notebooks Work About

written: 21 Apr 2026
rests on: 2
cited by: 5

The split type has to preserve the structure you're generalizing along

The chosen split type prior to training must mimic the split between “the known” and “the unknown” at the time of inference. Otherwise, we’re measuring the model on a task it was never trained for.

Rests on

The split unit follows what you generalize to
Validation and Test sets must have identical split structure to support the same claim a constraint when splitting

Cited by

Random splitting temporal data makes the target interpolation not forecasting a general rule when deciding on the split type
Test set's split structure determines the claim about the model a general rule on splitting datasets
The split unit follows what you generalize to a more universal rule.
Validation and Test sets must have identical split structure to support the same claim general principle this is an instance of.
Validation set's split structure selects the model a general rule on splitting datasets