PSO/benchmark_results/pso_v4_deep_accuracy.csv at 89850f592e5fcc4a3dfe3a25f40eca05d48ba052

mirror of https://github.com/jung-geun/PSO.git synced 2026-09-20 14:11:48 +09:00

Files

T

jung-geun 813433000a feat: modernize PSO and add convergence research

Migrate the package and examples to the tensor-native PyTorch implementation, add benchmark evidence, and add the guarded post-training convergence protocol with TensorBoard progress monitoring and hash-verified recovery.

Constraint: Preserve one-shot official-test sealing and auditable research artifacts

Rejected: Commit local .omc runs and downloaded datasets | multi-gigabyte runtime state is machine-local

Confidence: high

Scope-risk: broad

Not-tested: Production CUDA run on pieroot-server

2026-09-07 22:03:25 +09:00

2.5 KiB

Raw Blame History

1	lane	profile_or_arch	seed	model_name	param_count	initial_test_acc	final_test_acc	final_test_loss	subset_fitness_acc	subset_fitness_loss	pso_epochs	adam_epochs	elapsed_sec	model_fingerprint	data_fingerprint
2	architecture	raw_linear	101	Raw Linear (784->10)	7850	0.1449	0.9239	0.269691			0	10	2.6001	12cef21a80b85b0b	8dd702555745641a
3	architecture	raw_linear	102	Raw Linear (784->10)	7850	0.1308	0.924	0.270518			0	10	2.6265	0f099d5ae79e7863	8dd702555745641a
4	architecture	raw_linear	103	Raw Linear (784->10)	7850	0.0559	0.9257	0.266626			0	10	2.8535	23a4a6362abdad3d	8dd702555745641a
5	architecture	raw_mlp	101	Raw MLP (784->128->64->10)	109386	0.0812	0.9773	0.076661			0	10	8.8041	9ac96d5f45d71cb3	8dd702555745641a
6	architecture	raw_mlp	102	Raw MLP (784->128->64->10)	109386	0.0522	0.976	0.081621			0	10	5.7141	55966246db4dfff1	8dd702555745641a
7	architecture	raw_mlp	103	Raw MLP (784->128->64->10)	109386	0.1111	0.9776	0.076823			0	10	4.5491	723217e2418edcfd	8dd702555745641a
8	architecture	compact_cnn	101	Compact CNN (9,098 params)	9098	0.0851	0.9866	0.039444			0	10	6.228	db41fb515dcb49fa	8dd702555745641a
9	architecture	compact_cnn	102	Compact CNN (9,098 params)	9098	0.0963	0.9858	0.042864			0	10	5.396	efd743cad60c7530	8dd702555745641a
10	architecture	compact_cnn	103	Compact CNN (9,098 params)	9098	0.1072	0.9836	0.04912			0	10	5.4935	17891ec08e79ee74	8dd702555745641a
11	optimizer	adam_only	101	Compact CNN (Adam-Only)	9098	0.0851	0.9866	0.039444			0	10	6.228	db41fb515dcb49fa	8dd702555745641a
12	optimizer	adam_only	102	Compact CNN (Adam-Only)	9098	0.0963	0.9858	0.042864			0	10	5.396	efd743cad60c7530	8dd702555745641a
13	optimizer	adam_only	103	Compact CNN (Adam-Only)	9098	0.1072	0.9836	0.04912			0	10	5.4935	17891ec08e79ee74	8dd702555745641a
14	optimizer	pso_only	101	Compact CNN (PSO-Only)	9098	0.0851	0.3901	22.182518	0.399	21.674469	40	0	2.5761	db41fb515dcb49fa	8dd702555745641a
15	optimizer	pso_only	102	Compact CNN (PSO-Only)	9098	0.0963	0.3886	6.031698	0.3925	5.977414	40	0	2.331	efd743cad60c7530	8dd702555745641a
16	optimizer	pso_only	103	Compact CNN (PSO-Only)	9098	0.1072	0.3242	14.099036	0.3405	14.065619	40	0	2.7469	17891ec08e79ee74	8dd702555745641a
17	optimizer	hybrid	101	Compact CNN (Hybrid)	9098	0.0851	0.9658	0.109308	0.399	21.674469	40	10	7.4414	db41fb515dcb49fa	8dd702555745641a
18	optimizer	hybrid	102	Compact CNN (Hybrid)	9098	0.0963	0.9807	0.061528	0.3925	5.977414	40	10	7.226	efd743cad60c7530	8dd702555745641a
19	optimizer	hybrid	103	Compact CNN (Hybrid)	9098	0.1072	0.9725	0.088789	0.3405	14.065619	40	10	7.7426	17891ec08e79ee74	8dd702555745641a

2.5 KiB Raw Blame History

2.5 KiB

Raw Blame History