Neuro-Symbolic Learning of Answer Set Programs from Raw Data

Authors: Daniel Cunnington, Mark Law, Jorge Lobo, Alessandra Russo

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

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate NSIL on three problem domains of different complexity, including an NP-complete problem. Our results demonstrate that NSIL learns expressive knowledge, solves computationally complex problems, and achieves state-of-the-art performance in terms of accuracy and data efficiency.
Researcher Affiliation Collaboration Daniel Cunnington1,2 , Mark Law3 , Jorge Lobo4 and Alessandra Russo2 1IBM Research Europe 2Imperial College London 3ILASP Limited 4ICREA-Universitat Pompeu Fabra dancunnington@uk.ibm.com, mark@ilasp.com, jorge.lobo@upf.edu, a.russo@imperial.ac.uk
Pseudocode Yes To summarise, NSIL solves Equations 1 and 2 with the following steps: 1. The symbolic learner learns an initial knowledge base H that satisfies B and covers each possible label in Y. The neural network parameters θ are initialised randomly. 2. The neural network is trained for 1 epoch on the training set D to update θ , using the ASP program Πy for each X, y D. Note that Πy is constructed using H . 3. The weights of the corrective examples are updated (see Section 3.3), and a new H is learned. 4. Steps 2-3 are repeated for a fixed number of iterations. NSIL outputs the learned neural network parameters θ and the learned knowledge base H.
Open Source Code Yes Code and technical appendix: https://github.com/Dan Cunnington/NSIL
Open Datasets Yes The dataset D contains X, y samples where X is a sequence of MNIST digit images, and y is the cumulative addition or product of the digits in X. ... In our experiments, we replace elements {1 .. 5} in U with corresponding MNIST images from classes 1-5 (i.e., Z = {1 .. 5}). We also present results for both variants with a more difficult perception task using Fashion MNIST images instead of MNIST.
Dataset Splits Yes The training set contains 3,000 samples of sequences of length 2-5, and three test sets with sequences of lengths 5, 10, and 100 in the Addition task, and 5, 10, and 15 in the Product task, to verify the trained network and learned knowledge can extrapolate to longer inputs. Each test set contains 10,000 samples. ... Both tasks use training, validation and test datasets of size 24,000, 6,000, and 5,000 samples respectively. ... Training, validation and test datasets contain 1502, 376, and 316 examples respectively.
Hardware Specification Yes Neural networks were trained on an NVIDIA A100 GPU
Software Dependencies No The paper mentions using 'PyTorch' but does not specify a version number or other software dependencies with versions.
Experiment Setup Yes We perform hyper-parameter tuning for all methods using a held-out validation set and a separate random seed. ... We trained the neural networks using Adam optimizer with an initial learning rate of 0.001. ... λ = 1 in all of our experiments.