Towards explicit representation of an artificial neural network model: Comparison of two artificial neural network rule extraction approaches
Bibliographic record
Abstract
In the quest for interpretable models, two versions of a neural network rule extraction algorithm were proposed and compared. The two algorithms are called the Piece-Wise Linear Artificial Neural Network (PWL-ANN) and enhanced Piece-Wise Linear Artificial Neural Network (enhanced PWL-ANN) algorithms. The PWL-ANN algorithm is a decomposition artificial neural network (ANN) rule extraction algorithm, and the enhanced PWL-ANN algorithm improves upon the PWL-ANN algorithm and extracts multiple linear regression equations from a trained ANN model by approximating the hidden sigmoid activation functions using N-piece linear equations. In doing so, the algorithm provides interpretable models from the originally trained opaque ANN models. A detailed application case study illustrates how the generated enhanced-PWL-ANN models can provide understandable IF-THEN rules about a problem domain. Comparison of the results generated by the two versions of the PWL-ANN algorithm showed that in comparison to the PWL-ANN models, the enhanced-PWL-ANN models support improved fidelities to the originally trained ANN models. The results also showed that more concise rule sets could be generated using the enhanced-PWL-ANN algorithm. If a more simplified set of rules is desired, the enhanced-PWL-ANN algorithm can be combined with the decision tree approach. Potential application of the algorithms to domains related to petroleum engineering can help enhance understanding of the problems.
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 | 0.003 | 0.008 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.002 | 0.003 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 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".