Can We Do Better with What We Have Done? Unveiling the Potential of ML Pipeline in Notebooks
Bibliographic record
Abstract
Computational notebooks are widely adopted by data scientists for experimenting with machine learning (ML) models. Despite the support for exploratory programming enabled by notebooks, they fall short in the ability to manage alternatives across different stages of the ML pipeline. In this study, we conduct a qualitative analysis to examine how data scientists explore various alternatives through a series of versions of notebooks on Kaggle. The findings indicate that data scientists investigate multiple alternatives at each stage across multiple versions, yet only a limited number of combinations from different stages are explored. Next, by combining alternatives from all stages to form previously unexplored paths, we discover that certain untested combinations of alternatives can outperform the best models as identified in the original notebooks. Moreover, by substituting the hyperparameter optimization and model configuration stages with AutoML methods, we observe that only a select number of ML pipelines experience improvement via AutoML, which implies the limitation of the current AutoML techniques. In summary, our study provides insights into the systematic and effective exploration of overlooked ML pipeline configuration combinations that yield superior results. The findings shed light on future research directions such as the development of tooling support of alternative management while striking a balance between manual exploration and automated optimization.
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 | 0.020 | 0.069 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.002 |
| Science and technology studies | 0.002 | 0.007 |
| Scholarly communication | 0.010 | 0.017 |
| Open science | 0.003 | 0.006 |
| Research integrity | 0.002 | 0.003 |
| Insufficient payload (model declined to judge) | 0.008 | 0.002 |
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".