Bibliographic record
Abstract
A sample compression scheme of size k for a concept class C is a pair of functions (f; g) called the compression function and the reconstruction function. The functions have the property that for any sample S consistent with a concept in C, f compresses S to some subset of S, for which g returns a set of domain points, labelled consistently with the original sample S. The sample compression scheme is called labelled if the compression sets are labelled subsets of S and unlabelled if the compression sets are subsets of the instance set of S. M. Warmuth and S. Floyd have shown that if a sample compression scheme of size equal to the VC dimension of a concept class C exists then C can be PAC learned by some learning algorithm. Although it is already known that any concept class of nite VC dimension is PAC learnable, the existence of a sample compression scheme of size equal to the VC dimension improves the sample complexity of learning some concept classes. This leads to an important conjecture, rst proposed by M. Warmuth and S. Floyd: does there always exist a sample compression scheme of size O(d) for a concept class C with VC dimension d. This thesis examines a modi cation of sample compression schemes, speci cally, for a concept class C we de ne a sequence-based sample compression scheme for C as a pair of functions (f ; g ) where the items we compress to are now sequences instead of sets. Here we can di erentiate between labelled and unlabelled sequence-based sam- ple compression schemes in a similar fashion as with standard sample compression schemes. We look at properties of sequence-based sample compression schemes and also discuss a few sequence-based sample compression scheme algorithms and deter- mine how they improve compression bounds over the original set-based compression scheme algorithms. Finally we discuss connections between set and sequence-based sample compression schemes and design theory. ii
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 | 0.004 | 0.024 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.002 | 0.002 |
| Bibliometrics | 0.004 | 0.006 |
| Science and technology studies | 0.003 | 0.007 |
| Scholarly communication | 0.010 | 0.019 |
| Open science | 0.003 | 0.007 |
| Research integrity | 0.002 | 0.006 |
| Insufficient payload (model declined to judge) | 0.018 | 0.003 |
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".