Pseudo-Independent Models and Decision Theoretic Knowledge Discovery
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Graphical models such as Bayesian networks (BNs) (Pearl, 1988; Jensen & Nielsen, 2007) and decomposable Markov networks (DMNs) (Xiang, Wong., & Cercone, 1997) have been widely applied to probabilistic reasoning in intelligent systems. Knowledge representation using such models for a simple problem domain is illustrated in Figure 1: Virus can damage computer files and so can a power glitch. Power glitch also causes a VCR to reset. Links and lack of them convey dependency and independency relations among these variables and the strength of each link is quantified by a probability distribution. The networks are useful for inferring whether the computer has virus after checking files and VCR. This chapter considers how to discover them from data. Discovery of graphical models (Neapolitan, 2004) by testing all alternatives is intractable. Hence, heuristic search are commonly applied (Cooper & Herskovits, 1992; Spirtes, Glymour, & Scheines, 1993; Lam & Bacchus, 1994; Heckerman, Geiger, & Chickering, 1995; Friedman, Geiger, & Goldszmidt, 1997; Xiang, Wong, & Cercone, 1997). All heuristics make simplifying assumptions about the unknown data-generating models. These assumptions preclude certain models to gain efficiency. Often assumptions and models they exclude are not explicitly stated. Users of such heuristics may suffer from such exclusion without even knowing. This chapter examines assumptions underlying common heuristics and their consequences to graphical model discovery. A decision theoretic strategy for choosing heuristics is introduced that can take into account a full range of consequences (including efficiency in discovery, efficiency in inference using the discovered model, and cost of inference with an incorrectly discovered model) and resolve the above issue.
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.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
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 it