Bibliographic record
Abstract
Converging evidence from the behavioral sciences suggests that inference in probabilistic generative models can capture how human and animal subjects solve a variety of cognitive tasks involving uncertainty. In fact, such evidence has accumulated more quickly than bespoke computational modeling can keep pace, contributing impetus to the development of probabilistic programming systems to reduce the duplication of effort. Among these, deep probabilistic programming systems enable the integration of deep neural networks into probabilistic generative models as powerful function approximators. In this dissertation, I provide compositional tools for reasoning about model structure and inference in the deep probabilistic programming system Probabilistic Torch. I begin with a case report about applying deep probabilistic programming to neuroimaging, including mean-field inference and a nontrivial stochastic gradient estimator for the training objective. I then dive into my work on training neural proposals to break down a complex joint distribution into a series of simpler complete conditional distributions. I then describe my contributions to a domain-specific language of inference combinators for composing importance sampling strategies in deep probabilistic programs, including an inference combinators implementation of the previous decompositional sampling strategy. I finally overview my work on providing categorical foundations in which to ground the semantics of deep probabilistic programs and reason compositionally about their model structure and inference strategies. I then close the dissertation by discussing how compositionality can help probabilistic machine learning scale up towards brain-sized inference problems.--Author's abstract
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 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) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 0.000 |
| 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 itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, 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".