Predicting sample size required for classification performance
Rosa L. Figueroa, Qing Zeng‐Treitler, Sasikiran Kandula, Long H. Ngo · 2012
Read Predicting sample size required for classification performance by Rosa L. Figueroa, Qing Zeng‐Treitler, Sasikiran Kandula, Long H. Ngo (2012) 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/openalex-w2132886902.txt.
Description
Rosa Figueroa, Qing Zeng-Treitler, Sasikiran Kandula, and Long Ngo published this study in BMC Medical Informatics and Decision Making in 2012. They address a planning problem in machine learning: labeled training data are scarce and costly. Researchers need to know how many labeled examples a classifier needs before they pay for the labeling.
The authors fit an inverse power law curve to early points on a learning curve. Their fitting weights points by their reliability, then predicts scores for larger sample sizes. They tested the method on clinical text and waveform tasks, fitting hundreds of models. The weighted method beat a plain unweighted fit, giving researchers a practical planning tool.
Catalog record
| CREATOR | Rosa L. Figueroa, Qing Zeng‐Treitler, Sasikiran Kandula, Long H. Ngo |
|---|---|
| YEAR | 2012 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 38 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | OpenAlex / BMC Medical Informatics and Decision Making |