On the Markov Property of Neural Algorithmic Reasoning: Analyses and Methods
Abstract: Neural algorithmic reasoning is an emerging research direction that endows neural networks with the ability to mimic algorithmic executions step-by-step. A common paradigm in existing designs involves the use of historical embeddings in predicting the results of future execution steps. Our observation in this work is that such historical dependence intrinsically contradicts the Markov nature of algorithmic reasoning tasks. Based on this motivation, we present our ForgetNet, which does not use historical embeddings and thus is consistent with the Markov nature of the tasks. To address challenges in training ForgetNet at early stages, we further introduce G-ForgetNet, which uses a gating mechanism to allow for the selective integration of historical embeddings. Such an enhanced capability provides valuable computational pathways during the model's early training phase. Our extensive experiments, based on the CLRS-30 algorithmic reasoning benchmark, demonstrate that both ForgetNet and G-ForgetNet achieve better generalization capability than existing methods. Furthermore, we investigate the behavior of the gating mechanism, highlighting its degree of alignment with our intuitions and its effectiveness for robust performance.
- Relational inductive biases, deep learning, and graph networks. arXiv preprint arXiv:1806.01261, 2018.
- Learning to configure computer networks with neural algorithmic reasoning. Advances in Neural Information Processing Systems, 35:730–742, 2022.
- Neural algorithmic reasoning with causal regularisation. In International Conference on Machine Learning, pp. 2272–2288. PMLR, 2023.
- Introduction to algorithms. MIT press, 2022.
- Neural algorithmic reasoners are implicit planners. Advances in Neural Information Processing Systems, 34:15529–15542, 2021.
- Relational attention: Generalizing transformers for graph-structured tasks. In The Eleventh International Conference on Learning Representations, 2023.
- Graph neural networks are dynamic programmers. Advances in Neural Information Processing Systems, 35:20635–20647, 2022.
- Benchmarking graph neural networks. Journal of Machine Learning Research, 24(43):1–48, 2023.
- Neural bipartite matching. arXiv preprint arXiv:2005.11304, 2020.
- Neural message passing for quantum chemistry. In International conference on machine learning, pp. 1263–1272. PMLR, 2017.
- Neural turing machines. arXiv preprint arXiv:1410.5401, 2014.
- Hybrid computing using a neural network with dynamic external memory. Nature, 538(7626):471–476, 2016.
- Relational inductive bias for physical construction in humans and machines. arXiv preprint arXiv:1806.01203, 2018.
- A generalist neural algorithmic learner. In Learning on Graphs Conference, pp. 2–1. PMLR, 2022.
- Inferring algorithmic patterns with stack-augmented recurrent nets. Advances in neural information processing systems, 28, 2015.
- Neural gpus learn algorithms. arXiv preprint arXiv:1511.08228, 2015.
- Adam: A method for stochastic optimization. In The Eleventh International Conference on Learning Representations, 2015.
- Towards better out-of-distribution generalization of neural algorithmic reasoning tasks. Transactions on Machine Learning Research, 2022.
- Dual algorithmic reasoning. In The Eleventh International Conference on Learning Representations, 2023.
- End-to-end memory networks. Advances in neural information processing systems, 28, 2015.
- Neural algorithmic reasoning. Patterns, 2(7), 2021.
- Graph attention networks. In International Conference on Learning Representations, 2018.
- Pointer graph networks. Advances in Neural Information Processing Systems, 33:2232–2244, 2020a.
- Neural execution of graph algorithms. In International Conference on Learning Representations, 2020b.
- The CLRS algorithmic reasoning benchmark. In International Conference on Machine Learning, pp. 22084–22102. PMLR, 2022a.
- Reasoning-modulated representations. In Learning on Graphs Conference, pp. 50–1. PMLR, 2022b.
- How to transfer algorithmic reasoning knowledge to learn new algorithms? Advances in Neural Information Processing Systems, 34:19500–19512, 2021.
- What can neural networks reason about? In International Conference on Learning Representations, 2020.
- Neural execution engines: Learning to execute subroutines. Advances in Neural Information Processing Systems, 33:17298–17308, 2020.
- Deep sets. Advances in neural information processing systems, 30, 2017.
- Learning to execute. arXiv preprint arXiv:1410.4615, 2014.
Paper Prompts
Sign up for free to create and run prompts on this paper using GPT-5.
Top Community Prompts
Collections
Sign up for free to add this paper to one or more collections.