Enabling Knowledge Refinement upon New Concepts in Abductive Learning

Authors: Yu-Xuan Huang, Wang-Zhou Dai, Yuan Jiang, Zhi-Hua Zhou

AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on three neuro-symbolic learning tasks verified the effectiveness of the proposed approach. All experiments are repeated ten times on a server with Intel Xeon Gold 6242R CPU and Nvidia RTX 3090 GPU.
Researcher Affiliation Academia National Key Laboratory for Novel Software Technology, Nanjing University, Nanjing 210023, China {huangyx, daiwz, jiangy, zhouzh}@lamda.nju.edu.cn
Pseudocode Yes Algorithm 1 ABLnc; Algorithm 2 Conflict Resolution; Algorithm 3 Matching Knowledge Graph
Open Source Code Yes The code is available for download1. 1https://github.com/AbductiveLearning/ABLnc
Open Datasets Yes The dataset consists of 10k pairs of images, where the digits are randomly generated and their images are randomly sampled from the training set of MNIST.
Dataset Splits Yes The hyperparameters of ABLnc are determined by cross-validation on training data.
Hardware Specification Yes All experiments are repeated ten times on a server with Intel Xeon Gold 6242R CPU and Nvidia RTX 3090 GPU.
Software Dependencies No The paper mentions software components such as CNN, LOF, ASP, ILASP, and Transformer, but does not provide specific version numbers for any of these dependencies.
Experiment Setup No The paper mentions that 'The hyperparameters of ABLnc are determined by cross-validation on training data' but does not provide specific values for these hyperparameters, training configurations, or other system-level settings.