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 (kaykenneth@gmail.com).

• Feedforward models
• ti254 — MLP, 100 seeds (regularization: 0.001)
• ti253 — LR, 100 seeds (regularization: 0.1)
• Basic-Delay RNN
• ti350 — tanh, delay=20, input_train=1, output_train=1, 200 instances "f-RNN"
• ti351 — tanh, delay=20, input_train=0, output_train=0, 200 instances "r-RNN"
• ti352 — tanh, delay=20, input_train=0, output_train=1, 200 instances "r-RNN" with feedforward outputs trainable
• Extended-Delay RNN
• ti353 — tanh, delay=60, input_train=1, output_train=1, 200 instances "f-RNN"
• ti354 — tanh, delay=60, input_train=0, output_train=0, 200 instances "r-RNN"
• Variable-Delay RNN
• ti360 — tanh, delay=60, input_train=1, output_train=1, jit=0.67, 200 instances "f-RNN"
• ti361 — tanh, delay=60, input_train=0, output_train=0, jit=0.67, 200 instances "r-RNN"
• Feedforward-trainable RNN ("ff-RNN")
• ti365 — tanh, delay=20, input_train=1, output_train=1, 200 seeds, 500 instances Basic-delay
• ti366 — tanh, delay=60, input_train=1, output_train=1, 200 seeds, 500 instances Extended-delay
• ti367 — tanh, delay=60, input_train=1, output_train=1, jit=0.67, 200 seeds, 500 simulations Variable-delay