Studying how glomeruli can evolve from training.
Quick start using this jupyte notebook
Note: The following will be cleaned up shortly.
Figure 1: Standard network with a receptor layer
python main.py --analyze receptor_standard
Figure SX: Control for standard network without receptor layer
python main.py --analyze control_standard
Figure SX: Impact of pruning
python main.py --analyze control_pn2kc_prune_boolean
Figure SX: Impact of relabel dataset
python main.py --analyze control_relabel_prune
Figure SX: Training ORN-output network on relabel datasets
python main.py --analyze control_relabel_singlelayer
Figure SX: RNN
python main.py --analyze rnn_relabel
Figure SX: Vary correlation level of ORNs
python main.py --analyze vary_orn_corr_relabel
Figure SX: Impact of KC normalization
python main.py --analyze kc_norm
Figure 2: Multi-task training
python main.py --analyze multihead_standard
Figure 3: Meta learning
python main.py --analyze meta_standard
Figure SX: Meta learning controls
python main.py --analyze meta_control_standard
Figure 3: Scaling K-N plot
python main.py --analyze scaling
Figure SX: Vary number of ORs
python main.py --analyze vary_or
python main.py --analyze meta_vary_or
Figure 4: Effect of non-negativity
python main.py --analyze control_nonnegative
Figure 4: Vary number of KCs
python main.py --analyze control_vary_kc
Figure 4: Vary number of PNs
python main.py --analyze control_vary_pn
Figure 4: PN normalization
python main.py --analyze pn_norm_relabel