Error Consistency for Machine Learning Evaluation and Validation with Application to Biomedical Diagnostics
Bibliographic record
Abstract
Supervised machine learning classification is the most common example of artificial intelligence (AI) in industry and in academic research. These technologies predict whether a series of measurements belong to one of multiple groups of examples on which the machine was previously trained. Prior to real-world deployment, all implementations need to be carefully evaluated with hold-out validation, where the algorithm is tested on different samples than it was provided for training, in order to ensure the generalizability and reliability of AI models. However, established methods for performing hold-out validation do not assess the consistency of the mistakes that the AI model makes during hold-out validation. Here, we show that in addition to standard methods, an enhanced technique for performing hold-out validation-that also assesses the consistency of the sample-wise mistakes made by the learning algorithm-can assist in the evaluation and design of reliable and predictable AI models. The technique can be applied to the validation of any supervised learning classification application, and we demonstrate the use of the technique on a variety of example biomedical diagnostic applications, which help illustrate the importance of producing reliable AI models. The validation software created is made publicly available, assisting anyone developing AI models for any supervised classification application in the creation of more reliable and predictable technologies.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.135 | 0.386 |
| Meta-epidemiology (narrow) | 0.003 | 0.001 |
| Meta-epidemiology (broad) | 0.002 | 0.003 |
| Bibliometrics | 0.005 | 0.003 |
| Science and technology studies | 0.002 | 0.005 |
| Scholarly communication | 0.004 | 0.004 |
| Open science | 0.004 | 0.006 |
| Research integrity | 0.005 | 0.007 |
| Insufficient payload (model declined to judge) | 0.003 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".