Skip to content

README.md

Latest commit

 

History

History
111 lines (88 loc) · 27.8 KB

README.md

File metadata and controls

111 lines (88 loc) · 27.8 KB

Model training argument files

DeepPhiPLNN Synthetic binary decision landscapes

argfile training data nepochs patience batch size phi layers ncells sigma loss solver dt0 dt scheduling learning rate optimizer
model_phi1_1a_v_klv1 training data
validation data		 2000 100 250 16 32 32 16 200 0.05 kl heun 1e-1 stepped
bounds: [200 500 1000]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_phi1_1a_v_klv2 training data
validation data		 2000 100 250 16 32 32 16 200 0.05 klv2 heun 1e-1 stepped
bounds: [200 500 1000]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_phi1_1a_v_mmd1 training data
validation data		 2000 100 250 16 32 32 16 200 0.05 mmd
(multiscale, 0.2 0.5 0.9 1.3)		 heun 1e-1 stepped
bounds: [200 500 1000]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_phi1_1b_v_mmd1 training data
validation data		 2000 100 250 16 32 32 16 200 0.05 mmd
(multiscale, 0.2 0.5 0.9 1.3)		 heun 1e-1 stepped
bounds: [200 500 1000]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_phi1_1c_v_mmd1 training data
validation data		 2000 100 250 16 32 32 16 200 0.05 mmd
(multiscale, 0.2 0.5 0.9 1.3)		 heun 1e-1 stepped
bounds: [200 500 1000]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_phi1_2a_v_mmd1 training data
validation data		 2000 100 250 16 32 32 16 200 0.05 mmd
(multiscale, 0.2 0.5 0.9 1.3)		 heun 1e-1 stepped
bounds: [200 500 1000]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_phi1_2b_v_mmd1 training data
validation data		 2000 100 250 16 32 32 16 200 0.05 mmd
(multiscale, 0.2 0.5 0.9 1.3)		 heun 1e-1 stepped
bounds: [200 500 1000]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_phi1_2c_v_mmd1 training data
validation data		 2000 100 250 16 32 32 16 200 0.05 mmd
(multiscale, 0.2 0.5 0.9 1.3)		 heun 1e-1 stepped
bounds: [200 500 1000]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_phi1_3a_v_mmd1 training data
validation data		 2000 100 250 16 32 32 16 200 0.05 mmd
(multiscale, 0.2 0.5 0.9 1.3)		 heun 1e-1 stepped
bounds: [200 500 1000]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_phi1_3b_v_mmd1 training data
validation data		 2000 100 250 16 32 32 16 200 0.05 mmd
(multiscale, 0.2 0.5 0.9 1.3)		 heun 1e-1 stepped
bounds: [200 500 1000]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_phi1_3c_v_mmd1 training data
validation data		 2000 100 250 16 32 32 16 200 0.05 mmd
(multiscale, 0.2 0.5 0.9 1.3)		 heun 1e-1 stepped
bounds: [200 500 1000]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_phi1_4a_v_mmd1 training data
validation data		 2000 100 250 16 32 32 16 200 0.05 mmd
(multiscale, 0.2 0.5 0.9 1.3)		 heun 1e-1 stepped
bounds: [200 500 1000]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_phi1_4a_v_mmd2 training data
validation data		 2000 100 250 16 32 32 16 200 0.05 mmd
(multiscale, 0.2 0.5 0.9 1.3)		 heun 1e-1 stepped
bounds: [200 500 1000]
scales: [0.5 0.5 0.5]		 constant rms
m=0.5
decay=0.9
clip=1.0
model_phi1_4b_v_mmd1 training data
validation data		 2000 100 250 16 32 32 16 200 0.05 mmd
(multiscale, 0.2 0.5 0.9 1.3)		 heun 1e-1 stepped
bounds: [200 500 1000]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_phi1_4b_v_mmd2 training data
validation data		 2000 100 250 16 32 32 16 200 0.05 mmd
(multiscale, 0.2 0.5 0.9 1.3)		 heun 1e-1 stepped
bounds: [200 500 1000]
scales: [0.5 0.5 0.5]		 constant rms
m=0.5
decay=0.9
clip=1.0
model_phi1_4c_v_mmd1 training data
validation data		 2000 100 250 16 32 32 16 200 0.05 mmd
(multiscale, 0.2 0.5 0.9 1.3)		 heun 1e-1 stepped
bounds: [200 500 1000]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_phi1_4c_v_mmd2 training data
validation data		 2000 100 250 16 32 32 16 200 0.05 mmd
(multiscale, 0.2 0.5 0.9 1.3)		 heun 1e-1 stepped
bounds: [200 500 1000]
scales: [0.5 0.5 0.5]		 constant rms
m=0.5
decay=0.9
clip=1.0
model_phi2_1a_v_mmd1 training data
validation data		 2000 100 250 16 32 32 16 200 0.05 mmd
(multiscale, 0.2 0.5 0.9 1.3)		 heun 1e-1 stepped
bounds: [200 500 1000]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_phi2_1b_v_mmd1 training data
validation data		 2000 100 250 16 32 32 16 200 0.05 mmd
(multiscale, 0.2 0.5 0.9 1.3)		 heun 1e-1 stepped
bounds: [200 500 1000]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_phi2_1c_v_mmd1 training data
validation data		 2000 100 250 16 32 32 16 200 0.05 mmd
(multiscale, 0.2 0.5 0.9 1.3)		 heun 1e-1 stepped
bounds: [200 500 1000]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0

DeepPhiPLNN Quadratic landscapes

argfile training data nepochs patience batch size phi layers ncells sigma loss solver dt0 dt scheduling learning rate optimizer
model_phiq_1a_v_klv1 training data
validation data		 1000 100 250 16 32 32 16 200 0.05 kl heun 1e-1 stepped
bounds: [100 250 500]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_phiq_1a_v_klv2 training data
validation data		 1000 100 250 16 32 32 16 200 0.05 klv2 heun 1e-1 stepped
bounds: [100 250 500]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_phiq_1a_v_mmd1 training data
validation data		 1000 100 250 16 32 32 16 200 0.05 mmd
(multiscale, 0.2 0.5 0.9 1.3)		 heun 1e-1 stepped
bounds: [100 250 500]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_phiq_2a_v_mmd1 training data
validation data		 1000 100 250 16 32 32 16 200 0.05 mmd
(multiscale, 0.2 0.5 0.9 1.3)		 heun 1e-1 stepped
bounds: [100 250 500]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0

AlgebraicPL Synthetic binary decision landscapes

argfile training data nepochs patience batch size phi layers ncells sigma loss solver dt0 dt scheduling learning rate optimizer
model_algphi1_1a_v_kl training data
validation data		 500 100 250 200 0.05 kl heun 1e-1 stepped
bounds: [50 100 250]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_algphi1_1a_v_klv2 training data
validation data		 500 100 250 200 0.05 klv2 heun 1e-1 stepped
bounds: [50 100 250]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_algphi1_1a_v_mmd1 training data
validation data		 500 100 250 200 0.05 mmd
(multiscale, 0.2 0.5 0.9 1.3)		 heun 1e-1 stepped
bounds: [50 100 250]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_algphi2_1a_v_kl training data
validation data		 500 100 250 200 0.05 kl heun 1e-1 stepped
bounds: [50 100 250]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_algphi2_1a_v_klv2 training data
validation data		 500 100 250 200 0.05 klv2 heun 1e-1 stepped
bounds: [50 100 250]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_algphi2_1a_v_mmd1 training data
validation data		 500 100 250 200 0.05 mmd
(multiscale, 0.2 0.5 0.9 1.3)		 heun 1e-1 stepped
bounds: [50 100 250]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0

AlgebraicPL Quadratic landscapes

argfile training data nepochs patience batch size phi layers ncells sigma loss solver dt0 dt scheduling learning rate optimizer
model_algphiq_1a_v_kl1 training data
validation data		 500 100 250 200 0.05 kl heun 1e-1 stepped
bounds: [50 100 250]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_algphiq_1a_v_klv21 training data
validation data		 500 100 250 200 0.05 klv2 heun 1e-1 stepped
bounds: [50 100 250]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_algphiq_1a_v_mmd1 training data
validation data		 500 100 250 200 0.05 mmd
(multiscale, 0.2 0.5 0.9 1.3)		 heun 1e-1 stepped
bounds: [50 100 250]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_algphiq_1a_v_mmd2 training data
validation data		 500 100 250 50 0.05 mmd
(multiscale, 0.2 0.5 0.9 1.3)		 heun 1e-1 stepped
bounds: [50 100 250]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_algphiq_1a_v_mmd3 training data
validation data		 500 100 250 200 0.05 mmd
(multiscale, 0.2 0.5 0.9 1.3)		 heun 1e-1 stepped
bounds: [50 100 250]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.0
decay=0.9
clip=1.0
model_algphiq_1a_v_mmd4 training data
validation data		 500 100 250 200 0.05 mmd
(multiscale, 0.1)		 heun 1e-1 stepped
bounds: [50 100 250]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.0
decay=0.9
clip=1.0
model_algphiq_1a_v_mmd5 training data
validation data		 500 100 250 200 0.05 mmd
(multiscale, 10)		 heun 1e-1 stepped
bounds: [50 100 250]
scales: [0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.0
decay=0.9
clip=1.0
model_algphiq_1a_v_mmd6 training data
validation data		 500 100 250 200 0.05 mmd
(multiscale, 10)		 heun 1e-1 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.0
decay=0.9
clip=1.0

FACS Landscapes

argfile training data nepochs patience batch size phi layers ncells sigma loss solver dt0 dt scheduling learning rate optimizer
model_facs_dec1_v1_argset1 training data
validation data		 1000 200 50 16 32 32 16 800 0.05 mmd
(multiscale, 0.2 0.5 0.9 1.1 1.3 1.5 2.0)		 heun 5e-3 stepped
bounds: [50 100 200 300]
scales: [0.5 0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_facs_dec1_v1_argset2 training data
validation data		 1000 200 50 16 32 32 16 800 0.05 mmd
(multiscale, 1.3547579050064087)		 heun 5e-3 stepped
bounds: [50 100 200 300]
scales: [0.5 0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_facs_dec1_v2_argset1 training data
validation data		 1000 200 50 16 32 32 16 800 0.05 mmd
(multiscale, 0.2 0.5 0.9 1.1 1.3 1.5 2.0)		 heun 5e-3 stepped
bounds: [50 100 200 300]
scales: [0.5 0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_facs_dec1_v2_argset2 training data
validation data		 1000 200 50 16 32 32 16 800 0.05 mmd
(multiscale, 0.05 0.10 0.15 0.5)		 heun 5e-3 stepped
bounds: [50 100 200 300]
scales: [0.5 0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_facs_dec1_v2_argset3 training data
validation data		 1000 200 50 16 32 32 32 32 16 800 0.05 mmd
(multiscale, 0.05 0.10 0.15 0.5)		 heun 5e-3 stepped
bounds: [50 100 200 300]
scales: [0.5 0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_facs_dec1_v2_argset4 training data
validation data		 1000 200 50 16 32 32 16 800 0.05 mmd
(multiscale, 0.23435935378074646)		 heun 5e-3 stepped
bounds: [50 100 200 300]
scales: [0.5 0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_facs_dec1_v3_argset1 training data
validation data		 1000 200 50 16 32 32 32 32 16 800 0.05 mmd
(multiscale, 0.05 0.10 0.15 0.5)		 heun 5e-3 stepped
bounds: [50 100 200 300]
scales: [0.5 0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_facs_dec1_v3_argset2 training data
validation data		 1000 200 50 16 32 32 32 32 16 800 0.05 mmd
(multiscale, 0.11244228482246399)		 heun 5e-3 stepped
bounds: [50 100 200 300]
scales: [0.5 0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_facs_dec1_v4_argset1 training data
validation data		 1000 200 50 16 32 32 16 800 0.05 mmd
(multiscale, 1.6738450527191162)		 heun 5e-3 stepped
bounds: [100 300 500 700]
scales: [0.5 0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-4, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_facs_dec1_v4_argset2 training data
validation data		 1000 200 50 16 32 32 16 800 0.05 mmd
(multiscale, 0.2 0.5 1.0 1.6738450527191162)		 heun 5e-3 stepped
bounds: [100 300 500 700]
scales: [0.5 0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_facs_dec2_v1_argset1 training data
validation data		 1000 200 50 16 32 32 16 800 0.05 mmd
(multiscale, 0.2 0.5 0.9 1.1 1.3 1.5 2.0)		 heun 5e-3 stepped
bounds: [50 100 200 300]
scales: [0.5 0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_facs_dec2_v1_argset2 training data
validation data		 1000 200 50 16 32 32 16 800 0.05 mmd
(multiscale, 1.275067687034607)		 heun 5e-3 stepped
bounds: [50 100 200 300]
scales: [0.5 0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_facs_dec2_v1_argset3 training data
validation data		 1000 200 50 16 32 32 16 800 0.05 mmd
(multiscale, 1.275067687034607)		 heun 5e-3 stepped
bounds: [100 300 500 700]
scales: [0.5 0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-4, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_facs_dec2_v1_argset4 training data
validation data		 1000 200 50 16 32 32 16 800 0.05 mmd
(multiscale, 0.2 0.5 1.275067687034607)		 heun 5e-3 stepped
bounds: [100 300 500 700]
scales: [0.5 0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-4, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_facs_dec2_v2_argset1 training data
validation data		 1000 200 50 16 32 32 16 800 0.05 mmd
(multiscale, 0.2 0.5 0.9 1.1 1.3 1.5 2.0)		 heun 5e-3 stepped
bounds: [50 100 200 300]
scales: [0.5 0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_facs_dec2_v2_argset2 training data
validation data		 1000 200 50 16 32 32 16 800 0.05 mmd
(multiscale, 0.05 0.10 0.15 0.5)		 heun 5e-3 stepped
bounds: [50 100 200 300]
scales: [0.5 0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_facs_dec2_v2_argset3 training data
validation data		 1000 200 50 16 32 32 32 32 16 800 0.05 mmd
(multiscale, 0.05 0.10 0.15 0.5)		 heun 5e-3 stepped
bounds: [50 100 200 300]
scales: [0.5 0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_facs_dec2_v2_argset4 training data
validation data		 1000 200 50 16 32 32 16 800 0.05 mmd
(multiscale, 0.2213098108768463)		 heun 5e-3 stepped
bounds: [50 100 200 300]
scales: [0.5 0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0
model_facs_dec2_v3_argset1 training data
validation data		 1000 200 50 16 32 32 32 32 16 800 0.05 mmd
(multiscale, 0.05 0.10 0.15 0.5)		 heun 5e-3 stepped
bounds: [50 100 200 300]
scales: [0.5 0.5 0.5 0.5]		 exponential_decay
(1e-2, 1e-5, 50)		 rms
m=0.5
decay=0.9
clip=1.0

Misc. Landscapes

argfile training data nepochs patience batch size phi layers ncells sigma loss solver dt0 dt scheduling learning rate optimizer