Inferring Generative Model Structure with Static Analysis

Authors: Paroma Varma, Bryan D. He, Payal Bajaj, Nishith Khandwala, Imon Banerjee, Daniel Rubin, Christopher Ré

NeurIPS 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Empirically, Coral matches or outperforms traditional structure learning approaches by up to 3.81 F1 points. Using Coral to model dependencies instead of assuming independence results in better performance than a fully supervised model by 3.07 accuracy points when heuristics are used to label radiology data without ground truth labels.
Researcher Affiliation Academia Paroma Varma1, Bryan He2, Payal Bajaj2, Nishith Khandwala2, Imon Banerjee3, Daniel Rubin3,4, Christopher Ré2 1Electrical Engineering, 2Computer Science, 3Biomedical Data Science, 4Radiology Stanford University {paroma,bryanhe,pabajaj,nishith,imonb,rubin}@stanford.edu, chrismre@cs.stanford.edu
Pseudocode Yes def create_primitives(image,segmentation): P = Primitive Set() P.area = get_area(segmentation) P.perimeter = get_perimeter(segmentation) P.intensity = np.sum(segmentation*image) P.ratio = P.intensity/P.perimeter return P
Open Source Code No The paper does not include any explicit statement or link indicating that the source code for the described methodology is publicly available.
Open Datasets Yes Visual Genome and Activity Net Classification We explore how to extract complex relations in images and videos given object labels and their bounding boxes. We used subsets of two datasets, Visual Genome [21] and Activity Net [9]... Mammogram Tumor Classification We used the DDSM-CBIS [32] dataset, which consists of 1800 scanned film mammograms and associated segmentations for the tumors in the form of binary masks.
Dataset Splits No The paper does not explicitly provide specific dataset split information (percentages, sample counts, or clear predefined splits) for training, validation, and testing. While it mentions 'a small held out subset' for tuning in the bone tumor experiment, it lacks detailed, reproducible split information across all experiments.
Hardware Specification No The paper does not provide specific hardware details (such as exact GPU/CPU models, processor types, or memory amounts) used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details, such as library or solver names with version numbers, needed to replicate the experiments.
Experiment Setup No The paper describes the general experimental setup and model components but does not provide specific details such as concrete hyperparameter values (e.g., learning rate, batch size, number of epochs) or other system-level training settings needed for full reproducibility.