- Title
- Topological Dynamics of Functional Neural Network Graphs During Reinforcement Learning
- Creator
- Muller, Matthew; Kroon, Steve; Chalup, Stephan
- Relation
- ICONIP 2023: The 30th International Conference on Neural Information Processing. Neural Information Processing 30th International Conference, ICONIP 2023 Changsha, China, November 20–23, 2023 Proceedings, Part IX (Changsha, China 20-23 November, 2023) p. 190-204
- Relation
- ARC.DP210103304 http://purl.org/au-research/grants/arc/DP210103304
- Publisher Link
- http://dx.doi.org/10.1007/978-981-99-8138-0_16
- Publisher
- Springer
- Resource Type
- conference paper
- Date
- 2024
- Description
- This study investigates the topological structures of neural network activation graphs, with a focus on detecting higher-order Betti numbers during reinforcement learning. The paper presents visualisations of the neurotopological dynamics of reinforcement learning agents both during and after training, which are useful for different dynamics analyses which we explore in this work. Two applications are considered: frame-by-frame analysis of agent neurotopology and tracking per-neuron presence in cavity boundaries over training steps. The experimental analysis suggests that higher-order Betti numbers found in a neural network’s functional graph can be associated with learning more complex behaviours.
- Subject
- machine learning; neural information processing; Betti numbers; agent neurotopology
- Identifier
- http://hdl.handle.net/1959.13/1499106
- Identifier
- uon:54620
- Identifier
- ISBN:9789819981373
- Identifier
- ISSN:1865-0929
- Language
- eng
- Reviewed
- Hits: 1883
- Visitors: 1859
- Downloads: 0
| Thumbnail | File | Description | Size | Format |
|---|