PSO/benchmark_results/pso_v6_phase_b.csv at 6fe31394b9da8d89dff40c19974272b6852a28e9

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

1.1 KiB

Raw Blame History

1	protocol_version,MNIST-PSO-RAW-V6 1.0.0
2	--- Phase B Screen Results ---
3	config_id,description,val_nll,val_acc_%,queries,sample_evals,wall_time_sec
4	G0,exact V5 control,1.183117,67.73,4800,9600000,8.4356
5	G1,isolate anisotropic per-tensor scaling,1.437604,62.32,4800,9600000,6.6136
6	G2,isolate nonzero launch velocity,1.296105,60.72,4800,9600000,6.6528
7	G3,isolate mutation,1.130839,68.07,4800,9600000,6.4943
8	G4,velocity x mutation interaction,1.11628,65.76,4800,9600000,6.4837
9	G5,test sufficient bound expansion,0.90151,70.93,4800,9600000,6.76
10	G6,test broader normalized initialization,0.96129,69.56,4800,9600000,6.7408
11	G7,isolate antithetic position coupling against G2,1.291266,58.79,4800,9600000,6.4273
12	G8,retained semantic control (public Optimizer),0.883557,70.99,4800,9600000,8.7632
13	--- Phase B Confirmation Aggregates ---
14	config_id,mean_val_acc_%,std_val_acc,mean_val_nll,std_val_nll,num_seeds
15	G0,79.526667,0.576397,0.689687,0.008079,3
16	G1,76.953333,2.490067,0.833759,0.123667,3
17	G5,84.236667,0.903678,0.510297,0.024899,3
18	G6,84.25,0.219317,0.503809,0.015789,3
19	G8,84.916667,0.988804,0.48144,0.029447,3

1.1 KiB Raw Blame History

1.1 KiB

Raw Blame History