Emergent neural dynamics and geometry for generalization in a transitive inference task

Relational cognition, the ability to infer relationships that generalize to novel combinations of objects, is fundamental to human and animal intelligence. Despite this importance, it remains unclear how relational cognition is implemented in the brain due in part to a lack of hypotheses and predictions at the levels of collective neural activity and behavior. Here we discovered, analyzed, and experimentally tested neural networks (NNs) that perform transitive inference (TI), a classic relational task (if A > B and B > C, then A > C). We found NNs that (i) generalized perfectly, despite lacking overt transitive structure before training, (ii) generalized when the task required working memory (WM), a capacity thought to be essential to inference in the brain, (iii) emergently expressed behaviors long observed in living subjects, in addition to a novel order-dependent behavior, and (iv) adopted different task solutions yielding alternative behavioral and neural predictions. Furth..., Neural network (and other) models trained and analyzed on a transitive inference task (\"delay TI\"), as described in the study \"Emergent neural dynamics and geometry for generalization in a transitive inference task\".
Models were defined and trained using PyTorch., , # Emergent neural dynamics and geometry for generalization in a transitive inference task

[https://doi.org/10.5061/dryad.83bk3jb0v](https://doi.org/10.5061/dryad.83bk3jb0v)

Python (pickle) files containing PyTorch-based models trained on a transitive inference task (\"delay TI\"). These models are described and analyzed in the study \"Emergent neural dynamics and geometry in a transitive inference task\" (available as a preprint).

These models are readable using Jupyter Notebooks and code that has been deposited on Zenodo and is also available on GitHub (link in the Related Works section).

Below is the list of model types and their corresponding directories, each directory (titled \"ti#\") corresponding to a batch of model instances. Each file in the directory corresponds to a single model instance and contains the model parameters and additional information from training.

For any clarifications or requests please contact Kenneth Kay ([[e-mail])](mailto:kaykenneth@gmail.c).

...