Bibliographic record
Abstract
Imitation learning is a supervised learning problem that involves training a model to perform a task in a given environment using demonstrations of an expert.In this thesis, we propose 5 metrics to evaluate the performance of imitation learning agents.We compare state-of-the-art imitation learning models to deep neural networks at imitating state-based and reactive behavior.To compare the imitation learning techniques, we use two partially observable domains: the continuous RoboCup domain and the discrete Vacuum Cleaner domain.We show how our proposed metrics provide us with more qualitative information about the performance of imitation learners when imitating state-based behavior compared to state-of-the-art metrics.In addition, we show how our testing methodology provides results that resemble the eye-test that current testing methodologies fail to provide.We also show how Long Short-Term Memory (LSTM) networks outperform state-of-the-art models at imitating state-based behavior in the RoboCup soccer domain.i List of Tables 3.1 The experts used in the model of the vacuum cleaner domain that includes dirt.Those experts were used in the work of Ontañón et al. [1] and some of them were used in the work Gunaratne et al. [2] . . . . .26 3.2 The expert used in the study of Gunaratne et al. [2] This expert was used in the vacuum cleaner with dirt model. . . . . . . . . . . . . . .27 3.3 The experts used in the work of Tîrnȃucȃ et al. [3] with the vacuum cleaner model that excludes dirt. . . . . . . . . . . . . . . . . . . . .29 3.4 The experts used in the study of Gunaratne et al. [4] using the discretized model of RoboCup soccer. . . . . . . . .
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 | 0.001 | 0.004 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.002 |
| Open science | 0.001 | 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".