Best holdout assessment is sufficient for cancer transcriptomic model selection
Jake Crawford, Maria Chikina, Casey S. Greene · 2024
Read Best holdout assessment is sufficient for cancer transcriptomic model selection by Jake Crawford, Maria Chikina, Casey S. Greene (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-11701843.txt.
Description
Jake Crawford, Maria Chikina, and Casey Greene tested in 2024 a piece of modeling folklore: that simpler gene signatures travel better to new data. They trained mutation-predicting models on cancer cell lines and on tumors, then swapped the test sets, and also held out whole cancer types.
Choosing models by their score on a held-out set alone picked winners as reliably as rules favoring simpler models. Heavy shrinkage did not generalize better for either plain or neural models. The practical moral: trust a clean holdout score, and do not pay for simplicity that buys no reach.
Catalog record
| CREATOR | Jake Crawford, Maria Chikina, Casey S. Greene |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 51 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |