i-do-ai/fci-neuron-tcns

FCI neuron TCNs

The trained networks behind the Functional Complexity Index values reported in

Dendritic morphology and synaptic nonlinearities enhance functional complexity in human cortical neurons Ido Aizenbud, Daniela Yoeli, David Beniaguev, Christiaan P. J. de Kock, Michael London, Idan Segev. PNAS 123(28), e2533168123 (2026).

Code: https://github.com/ido4848/FCI. Training data: i-do-ai/fci-neuron-simulations.

For each of the 24 neuron models: the three seeds of the depth-3, width-128 temporal convolutional network of the paper (neuron_tcn_d3_w128_fk54_lk12_sd0_5_bs180_lr0_01_wd1e_08_vlw0_02_t{0,1,2} internally), trained to predict the neuron's output spikes and somatic voltage from its synaptic input. FCI = log10(1000 (1 - AUC)) / log10(1000 (1 - 0.9)), from the AUC on the normalized test set (a subset of the test simulations whose mean output rate is exactly 1 Hz).

Layout

models/<model>/results.csv, final_results.pkl                        final AUCs of all 12 networks trained on the model (depths 1, 2, 3, 7 x 3 seeds)
models/<model>/d3_w128_t<seed>/args.pkl, final_results.pkl           the network's training arguments and final results
models/<model>/d3_w128_t<seed>/models/model_<epoch>_<steps>          the best checkpoint (torch.save)
models/<model>/d3_w128_t<seed>/models/model_*_light.pkl              every evaluation along training (best_save + eval_history, plain pickle)

<model> is the model's folder name under simulating_neurons/neuron_models/ in the code repository. A d3_w128_t<seed> folder has the layout train_neuron_tcn.py writes, so

python calculate_fci.py --neuron_tcn_folder models/<model>/d3_w128_t0 --use_normalized

prints the FCI of that seed's checkpoint. The checkpoint is a dict with model_state_dict, args, in_chans, best_save, eval_history, epoch, step, ...; load it with Tcn.get_tcn_from_file(path, 'cpu') from training_nets/train_neuron_tcn.py. Compared with the checkpoints written during training, the optimizer state was removed and the weights were cast from float64 to float32 (355 MB → 59 MB per network; the network runs in float64 and its outputs change by ~1e-7). One checkpoint (Rat_L6_IPC_BBP_Mandge_diams_fixed_passive_dends_simple_soma, seed 1) was pruned before the release and only its light checkpoints, i.e. its AUC, survive.

Which number is in the paper

Every network was evaluated on the normalized test set every ~50,000 steps and its model_*_light.pkl files keep that history (eval_history['normalized_test'][step]['auc']). The checkpoint shipped for a seed is best_save: the evaluation with the highest normalized AUC over the whole run. The paper (Fig. 2A) took, per seed, the highest normalized AUC among the evaluations made within the first 500,000 training steps, and printed above each neuron the minimum FCI over its three seeds - the "Fig. 2A" column below, which reproduces the paper's 24 values exactly. Where a run kept improving after step 500,000 the shipped checkpoint is a little better than the paper's number for that seed.

import pickle, glob, numpy as np
from calculate_fci import auc_to_fci
light = max(glob.glob('models/<model>/d3_w128_t0/models/model_*_light.pkl'), key=lambda f: int(f.split('_')[-3]))
history = pickle.load(open(light, 'rb'))['eval_history']['normalized_test']
paper_auc = max(v['auc'] for step, v in history.items() if step <= 500000)
print(auc_to_fci(paper_auc))

Networks

Model	Seed	Checkpoint (step)	Normalized AUC	FCI	Paper: normalized AUC (step)	Paper: FCI	Fig. 2A
Rat_L2_TPC_BBP_Mandge_diams_fixed_passive_dends_simple_soma	d3_w128_t0	model_7_500040 (500,040)	0.99768	0.1830	0.99760 (300,300)	0.1902
Rat_L2_TPC_BBP_Mandge_diams_fixed_passive_dends_simple_soma	d3_w128_t1	model_6_450420 (450,420)	0.99763	0.1877	0.99763 (450,420)	0.1877	0.1877
Rat_L2_TPC_BBP_Mandge_diams_fixed_passive_dends_simple_soma	d3_w128_t2	model_7_500160 (500,160)	0.99740	0.2078	0.99728 (450,420)	0.2174
Rat_L23_PC_cADpyr229_1_BBP_passive_dends_simple_soma	d3_w128_t0	model_6_450495 (450,495)	0.99717	0.2260	0.99717 (450,495)	0.2260	0.2260
Rat_L23_PC_cADpyr229_1_BBP_passive_dends_simple_soma	d3_w128_t1	model_4_350415 (350,415)	0.99708	0.2327	0.99708 (350,415)	0.2327
Rat_L23_PC_cADpyr229_1_BBP_passive_dends_simple_soma	d3_w128_t2	model_7_500085 (500,085)	0.99729	0.2162	0.99706 (350,415)	0.2339
Rat_L23_PC_cADpyr229_5_BBP_passive_dends_simple_soma	d3_w128_t0	model_5_450450 (450,450)	0.99728	0.2170	0.99728 (450,450)	0.2170
Rat_L23_PC_cADpyr229_5_BBP_passive_dends_simple_soma	d3_w128_t1	model_5_450480 (450,480)	0.99730	0.2156	0.99730 (450,480)	0.2156	0.2156
Rat_L23_PC_cADpyr229_5_BBP_passive_dends_simple_soma	d3_w128_t2	model_8_500160 (500,160)	0.99705	0.2347	0.99686 (350,370)	0.2487
Rat_L4_TPC_BBP_Mandge_passive_dends_simple_soma	d3_w128_t0	model_3_300300 (300,300)	0.99745	0.2032	0.99745 (300,300)	0.2032	0.2032
Rat_L4_TPC_BBP_Mandge_passive_dends_simple_soma	d3_w128_t1	model_5_400380 (400,380)	0.99727	0.2177	0.99727 (400,380)	0.2177
Rat_L4_TPC_BBP_Mandge_passive_dends_simple_soma	d3_w128_t2	model_3_250200 (250,200)	0.99691	0.2452	0.99691 (250,200)	0.2452
Rat_L4_PC_cADpyr230_1_BBP_passive_dends_simple_soma	d3_w128_t0	model_5_400455 (400,455)	0.99674	0.2568	0.99674 (400,455)	0.2568
Rat_L4_PC_cADpyr230_1_BBP_passive_dends_simple_soma	d3_w128_t1	model_4_300330 (300,330)	0.99675	0.2560	0.99675 (300,330)	0.2560
Rat_L4_PC_cADpyr230_1_BBP_passive_dends_simple_soma	d3_w128_t2	model_9_500355 (500,355)	0.99719	0.2242	0.99703 (350,370)	0.2362	0.2362
Rat_L4_PC_cADpyr230_2_BBP_passive_dends_simple_soma	d3_w128_t0	model_5_400440 (400,440)	0.99673	0.2570	0.99673 (400,440)	0.2570
Rat_L4_PC_cADpyr230_2_BBP_passive_dends_simple_soma	d3_w128_t1	model_7_500160 (500,160)	0.99714	0.2283	0.99675 (300,480)	0.2560
Rat_L4_PC_cADpyr230_2_BBP_passive_dends_simple_soma	d3_w128_t2	model_7_500160 (500,160)	0.99705	0.2348	0.99691 (300,360)	0.2452	0.2452
Rat_L5_TPC_BBP_Mandge_passive_dends_simple_soma	d3_w128_t0	model_5_400320 (400,320)	0.99574	0.3148	0.99574 (400,320)	0.3148
Rat_L5_TPC_BBP_Mandge_passive_dends_simple_soma	d3_w128_t1	model_5_400320 (400,320)	0.99654	0.2697	0.99654 (400,320)	0.2697
Rat_L5_TPC_BBP_Mandge_passive_dends_simple_soma	d3_w128_t2	model_5_450360 (450,360)	0.99682	0.2510	0.99682 (450,360)	0.2510	0.2510
Rat_L5_TTPC1_cADpyr232_1_BBP_diams_fixed_passive_dends_simple_soma	d3_w128_t0	model_4_350280 (350,280)	0.99627	0.2857	0.99627 (350,280)	0.2857
Rat_L5_TTPC1_cADpyr232_1_BBP_diams_fixed_passive_dends_simple_soma	d3_w128_t1	model_5_450360 (450,360)	0.99682	0.2509	0.99682 (450,360)	0.2509	0.2509
Rat_L5_TTPC1_cADpyr232_1_BBP_diams_fixed_passive_dends_simple_soma	d3_w128_t2	model_7_500040 (500,040)	0.99622	0.2886	0.99618 (450,360)	0.2910
Rat_L5b_PC_2_Hay_passive_dends_simple_soma	d3_w128_t0	model_5_400365 (400,365)	0.99634	0.2815	0.99634 (400,365)	0.2815
Rat_L5b_PC_2_Hay_passive_dends_simple_soma	d3_w128_t1	model_6_450405 (450,405)	0.99706	0.2342	0.99706 (450,405)	0.2342	0.2342
Rat_L5b_PC_2_Hay_passive_dends_simple_soma	d3_w128_t2	model_7_500100 (500,100)	0.99709	0.2322	0.99662 (400,365)	0.2643
Rat_L6_UPC_BBP_Mandge_passive_dends_simple_soma	d3_w128_t0	model_7_500175 (500,175)	0.99661	0.2652	0.99629 (450,495)	0.2845
Rat_L6_UPC_BBP_Mandge_passive_dends_simple_soma	d3_w128_t1	model_7_500040 (500,040)	0.99702	0.2374	0.99698 (450,540)	0.2399	0.2399
Rat_L6_UPC_BBP_Mandge_passive_dends_simple_soma	d3_w128_t2	model_8_500175 (500,175)	0.99666	0.2618	0.99614 (250,245)	0.2930
Rat_L6_TPC_BBP_Mandge_passive_dends_simple_soma	d3_w128_t0	model_3_300360 (300,360)	0.99685	0.2495	0.99685 (300,360)	0.2495
Rat_L6_TPC_BBP_Mandge_passive_dends_simple_soma	d3_w128_t1	model_9_500280 (500,280)	0.99720	0.2237	0.99717 (300,360)	0.2260
Rat_L6_TPC_BBP_Mandge_passive_dends_simple_soma	d3_w128_t2	model_5_450540 (450,540)	0.99743	0.2052	0.99743 (450,540)	0.2052	0.2052
Rat_L6_IPC_BBP_Mandge_diams_fixed_passive_dends_simple_soma	d3_w128_t0	model_5_450600 (450,600)	0.99739	0.2081	0.99739 (450,600)	0.2081
Rat_L6_IPC_BBP_Mandge_diams_fixed_passive_dends_simple_soma	d3_w128_t1	light checkpoints only (450,540)	0.99752	0.1972	0.99752 (450,540)	0.1972	0.1972
Rat_L6_IPC_BBP_Mandge_diams_fixed_passive_dends_simple_soma	d3_w128_t2	model_5_400440 (400,440)	0.99744	0.2039	0.99744 (400,440)	0.2039
Human_L23_PC_0603_11_937_Eyal_passive_dends_simple_soma	d3_w128_t0	model_7_500160 (500,160)	0.99282	0.4280	0.99275 (450,300)	0.4303
Human_L23_PC_0603_11_937_Eyal_passive_dends_simple_soma	d3_w128_t1	model_3_250140 (250,140)	0.99277	0.4294	0.99277 (250,140)	0.4294	0.4294
Human_L23_PC_0603_11_937_Eyal_passive_dends_simple_soma	d3_w128_t2	model_7_500130 (500,130)	0.99286	0.4269	0.99234 (300,210)	0.4422
Human_L23_PC_1303_03_448_Eyal_passive_dends_simple_soma	d3_w128_t0	model_6_451035 (451,035)	0.99319	0.4165	0.99319 (451,035)	0.4165	0.4165
Human_L23_PC_1303_03_448_Eyal_passive_dends_simple_soma	d3_w128_t1	model_5_450495 (450,495)	0.99301	0.4222	0.99301 (450,495)	0.4222
Human_L23_PC_1303_03_448_Eyal_passive_dends_simple_soma	d3_w128_t2	model_3_250335 (250,335)	0.99267	0.4326	0.99267 (250,335)	0.4326
Human_L3_PC_0_BBP_passive_dends_simple_soma	d3_w128_t0	model_7_500040 (500,040)	0.99342	0.4092	0.99311 (400,560)	0.4190	0.4190
Human_L3_PC_0_BBP_passive_dends_simple_soma	d3_w128_t1	model_5_400560 (400,560)	0.99268	0.4324	0.99268 (400,560)	0.4324
Human_L3_PC_0_BBP_passive_dends_simple_soma	d3_w128_t2	model_7_500160 (500,160)	0.99257	0.4354	0.99202 (350,520)	0.4510
Human_L4_PC_BBP_Mandge_diams_fixed_passive_dends_simple_soma	d3_w128_t0	model_7_500085 (500,085)	0.99383	0.3951	0.99326 (300,330)	0.4143
Human_L4_PC_BBP_Mandge_diams_fixed_passive_dends_simple_soma	d3_w128_t1	model_5_450405 (450,405)	0.99381	0.3957	0.99381 (450,405)	0.3957	0.3957
Human_L4_PC_BBP_Mandge_diams_fixed_passive_dends_simple_soma	d3_w128_t2	model_6_450540 (450,540)	0.99323	0.4154	0.99323 (450,540)	0.4154
Human_L4_PC_539661667_Allen_passive_dends_simple_soma	d3_w128_t0	model_4_350250 (350,250)	0.99451	0.3697	0.99451 (350,250)	0.3697
Human_L4_PC_539661667_Allen_passive_dends_simple_soma	d3_w128_t1	model_5_450330 (450,330)	0.99469	0.3626	0.99469 (450,330)	0.3626	0.3626
Human_L4_PC_539661667_Allen_passive_dends_simple_soma	d3_w128_t2	model_4_350250 (350,250)	0.99428	0.3786	0.99428 (350,250)	0.3786
Human_L4_PC_569818704_Allen_passive_dends_simple_soma	d3_w128_t0	model_7_500175 (500,175)	0.99438	0.3747	0.99436 (450,495)	0.3757	0.3757
Human_L4_PC_569818704_Allen_passive_dends_simple_soma	d3_w128_t1	model_5_400320 (400,320)	0.99378	0.3969	0.99378 (400,320)	0.3969
Human_L4_PC_569818704_Allen_passive_dends_simple_soma	d3_w128_t2	model_7_500040 (500,040)	0.99395	0.3909	0.99393 (300,375)	0.3917
Human_L5_PC_BBP_Mandge_passive_dends_simple_soma	d3_w128_t0	model_7_500040 (500,040)	0.99494	0.3521	0.99458 (400,500)	0.3672
Human_L5_PC_BBP_Mandge_passive_dends_simple_soma	d3_w128_t1	model_4_350370 (350,370)	0.99419	0.3822	0.99419 (350,370)	0.3822
Human_L5_PC_BBP_Mandge_passive_dends_simple_soma	d3_w128_t2	model_5_450495 (450,495)	0.99471	0.3618	0.99471 (450,495)	0.3618	0.3618
Human_L5_PC_0_BBP_passive_dends_simple_soma	d3_w128_t0	model_6_450420 (450,420)	0.99455	0.3683	0.99455 (450,420)	0.3683
Human_L5_PC_0_BBP_passive_dends_simple_soma	d3_w128_t1	model_5_450405 (450,405)	0.99458	0.3672	0.99458 (450,405)	0.3672	0.3672
Human_L5_PC_0_BBP_passive_dends_simple_soma	d3_w128_t2	model_4_350310 (350,310)	0.99418	0.3826	0.99418 (350,310)	0.3826
Human_L5_PC_790872626_Allen_passive_dends_simple_soma	d3_w128_t0	model_5_450405 (450,405)	0.99333	0.4121	0.99333 (450,405)	0.4121
Human_L5_PC_790872626_Allen_passive_dends_simple_soma	d3_w128_t1	model_7_500085 (500,085)	0.99354	0.4051	0.99349 (450,405)	0.4068
Human_L5_PC_790872626_Allen_passive_dends_simple_soma	d3_w128_t2	model_5_400350 (400,350)	0.99388	0.3934	0.99388 (400,350)	0.3934	0.3934
Human_L6_PC_558211203_Allen_passive_dends_simple_soma	d3_w128_t0	model_7_500085 (500,085)	0.99307	0.4205	0.99275 (350,415)	0.4302
Human_L6_PC_558211203_Allen_passive_dends_simple_soma	d3_w128_t1	model_7_500175 (500,175)	0.99373	0.3988	0.99343 (400,860)	0.4088
Human_L6_PC_558211203_Allen_passive_dends_simple_soma	d3_w128_t2	model_6_450765 (450,765)	0.99368	0.4004	0.99368 (450,765)	0.4004	0.4004
Human_L6_PC_548494556_Allen_passive_dends_simple_soma	d3_w128_t0	model_5_450495 (450,495)	0.99574	0.3146	0.99574 (450,495)	0.3146	0.3146
Human_L6_PC_548494556_Allen_passive_dends_simple_soma	d3_w128_t1	model_5_400410 (400,410)	0.99505	0.3472	0.99505 (400,410)	0.3472
Human_L6_PC_548494556_Allen_passive_dends_simple_soma	d3_w128_t2	model_7_500175 (500,175)	0.99522	0.3396	0.99522 (450,495)	0.3399
Human_L6_PC_528614014_Allen_passive_dends_simple_soma	d3_w128_t0	model_7_500130 (500,130)	0.99553	0.3251	0.99488 (300,300)	0.3545
Human_L6_PC_528614014_Allen_passive_dends_simple_soma	d3_w128_t1	model_6_450390 (450,390)	0.99548	0.3274	0.99548 (450,390)	0.3274	0.3274
Human_L6_PC_528614014_Allen_passive_dends_simple_soma	d3_w128_t2	model_7_500130 (500,130)	0.99548	0.3274	0.99548 (350,340)	0.3277

Citation

@article{aizenbud2026fci,
    title   = {Dendritic morphology and synaptic nonlinearities enhance functional complexity in human cortical neurons},
    author  = {Aizenbud, Ido and Yoeli, Daniela and Beniaguev, David and de Kock, Christiaan P. J. and London, Michael and Segev, Idan},
    journal = {Proceedings of the National Academy of Sciences},
    volume  = {123},
    number  = {28},
    pages   = {e2533168123},
    year    = {2026},
    doi     = {10.1073/pnas.2533168123}
}