When training models, you would typically place your data into three buckets: Train, Test, and Hold Out. What is the purpose of

Question

FinnZ [79.3K]

2 years ago

14

When training models, you would typically place your data into three buckets: Train, Test, and Hold Out. What is the purpose of

having hold-out data

Mathematics

1 answer:

tekilochka [14] · Answer 1 · 2023-07-05T21:01:34+03:00

Answer:

So, having "hold-out" data allows you to have a nonpartisan set of data (so to speak) for you to refer to later, once you have made use of the "train" and "test" sets. That way you can make a fair assessment as to the difference between the new and old ('hold-out") sets.

Or if you need to repeat the experiment or model again - then you can refer to the "hold out data"