bgl3 · Fully connected

reference_bgl3_fc_modulof4_20261104-165730

Metrics

Set	mse	pearsonr	spearmanr	r2	above_wt_n	above_wt_spearmanr	budget_to_max	hit_rate_at_10	mean_at_10	max_at_10	captured_at_10	hit_rate_at_50	mean_at_50	max_at_50	precision_at_50	captured_at_50	hit_rate_at_100	mean_at_100	max_at_100	precision_at_100	captured_at_100	hit_rate_at_1pct	mean_at_1pct	max_at_1pct	precision_at_1pct	captured_at_1pct
train	1.9083	0.4369	0.4023	-0.0107	3798.0000	0.1003	6224.0000	0.7000	0.8247	2.2610	0.3445	0.6600	0.4897	2.2610	0.0000	0.2857	0.6600	0.5064	2.2610	0.0200	0.3093	0.6636	0.5050	2.2610	0.0187	0.3106
validation	1.5072	-0.0064	0.0109	-0.0006	338.0000	0.0522	223.0000	0.4000	-0.7796	1.0480	-0.0781	0.3200	-0.8601	2.0174	0.0200	-0.1251	0.2500	-0.8584	2.0174	0.0500	-0.1374	0.3846	-0.8615	1.0480	0.0000	-0.1101
test	1.3734	0.0552	0.0397	-0.0145	437.0000	0.0599	586.0000	0.4000	-0.0017	1.6946	0.1377	0.4600	-0.0417	2.2610	0.0200	0.1485	0.3600	-0.2856	2.2610	0.0900	0.0508	0.5625	0.3397	1.6946	0.0000	0.2707

Loss per epoch

training validation

Config

name: reference_bgl3_fc_modulof4
task: dms
data:
  dataset: bgl3
  encoding: one_hot,aa_index
  split_dir: null
  split_scheme: modulo
  split_fold: 4
model:
  name: fcn
  hidden:
  - 100
  - 100
  dropout: 0.2
train:
  epochs: 300
  batch_size: 32
  learning_rate: 0.0001
  patience: 15
  min_delta: 1.0e-05
  seed: 7