Learning Tractable Probabilistic Models from Inconsistent Local Estimates
Authors: Shasha Jin, Vasundhara Komaragiri, Tahrima Rahman, Vibhav Gogate
NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We show via experiments that our approach yields tractable models that are significantly superior to the ones learned from small amount of possibly noisy data, even when the local estimates are inconsistent. We performed a controlled empirical evaluation of our proposed method using 20 popular datasets that have been used in numerous studies on tractable models [10]. |
| Researcher Affiliation | Academia | Shasha Jin, Vasundhara Komaragiri, Tahrima Rahman, and Vibhav Gogate The University of Texas at Dallas |
| Pseudocode | Yes | Algorithm 1: LCN-LIS (X, E, {Pjk(Xj, Xk)|(Xj, Xk) E}, λ1, λ2, T) |
| Open Source Code | Yes | Did you include the code, data, and instructions needed to reproduce the main experimental results (either in the supplemental material or as a URL)? [Yes] In the Supplement |
| Open Datasets | Yes | We used 20 benchmark datasets that have been widely used in previous studies [20, 21] to evaluate our new approach. |
| Dataset Splits | Yes | We used 5-fold cross-validation to select the values of the hyperparameters λ1 and λ2. We used 10% of the randomly chosen examples in the training set to learn Q. |
| Hardware Specification | No | The paper does not provide specific details about the hardware used to run the experiments (e.g., GPU/CPU models, memory). |
| Software Dependencies | No | The paper mentions the use of certain algorithms and models but does not specify any software dependencies with version numbers. |
| Experiment Setup | Yes | We ran Algorithm 1 for a maximum of 48 hours or 1000 iterations (namely T = 1000) or convergence,whichever was earlier. We used 5-fold cross-validation to select the values of the hyperparameters λ1 and λ2. In steps 15–17, the algorithm updates the parameters using the gradient estimates gxi,ui and learning rate α. |