psaegert's picture
Publish catalogs (fastsrb, feynman, nguyen, v23-val, lample-charton-v23) v1
b5f283b verified
Raw
History Blame Contribute Delete
8.18 kB
metadata:
name: poly
version: 1
description: Poly suite incl. Poly-10.
sources:
- Poli 2003 lineage (via DSO aggregation)
- formulas + ranges from dso-org/deep-symbolic-optimization benchmarks.csv (BSD-3;
Petersen et al. 2021, Mundhenk et al. 2021)
sampling_defaults:
n_points: 20
method: random
noise: 0.0
source_kind: set
conventions:
sampling: vars use the `fastsrb` distribution (sample_range/sample_type); DSO
train_spec ranges; evenly-spaced (E) upstream grids are sampled uniformly here
(noted per entry as meta.dso_spec).
validity: 'accepted points follow the declared per-variable distribution CONDITIONED
on the expression''s valid domain; meta.finite_fraction is the MC-estimated
per-point valid fraction (n=20000); entries below 0.05 carry low_validity: true
and need per-entry review.'
expressions:
Poly-10:
raw: x1*x2+x3*x4+x5*x6+x1*x7*x9+x3*x6*x10
prepared: v1*v2+v3*v4+v5*v6+v1*v7*v9+v3*v6*v10
n_variables: 10
sources:
- Poli 2003 lineage (via DSO aggregation)
- formulas + ranges from dso-org/deep-symbolic-optimization benchmarks.csv (BSD-3;
Petersen et al. 2021, Mundhenk et al. 2021)
vars:
v1:
name: x1
sample_range:
- 0.0
- 1.0
sample_type:
- uni
- pos
v2:
name: x2
sample_range:
- 0.0
- 1.0
sample_type:
- uni
- pos
v3:
name: x3
sample_range:
- 0.0
- 1.0
sample_type:
- uni
- pos
v4:
name: x4
sample_range:
- 0.0
- 1.0
sample_type:
- uni
- pos
v5:
name: x5
sample_range:
- 0.0
- 1.0
sample_type:
- uni
- pos
v6:
name: x6
sample_range:
- 0.0
- 1.0
sample_type:
- uni
- pos
v7:
name: x7
sample_range:
- 0.0
- 1.0
sample_type:
- uni
- pos
v8:
name: x8
sample_range:
- 0.0
- 1.0
sample_type:
- uni
- pos
v9:
name: x9
sample_range:
- 0.0
- 1.0
sample_type:
- uni
- pos
v10:
name: x10
sample_range:
- 0.0
- 1.0
sample_type:
- uni
- pos
v0:
name: y
meta:
dso_spec:
train:
all:
U:
- 0
- 1
- 330
test:
all:
U:
- 0
- 1
- 170
function_set: None
finite_fraction: 1.0
Poly-1:
raw: x2/sqrt(pow(x1,2)+pow(x2,2)+pow(x3,2))
prepared: v2/sqrt(((v1)**(2))+((v2)**(2))+((v3)**(2)))
n_variables: 3
sources:
- Poli 2003 lineage (via DSO aggregation)
- formulas + ranges from dso-org/deep-symbolic-optimization benchmarks.csv (BSD-3;
Petersen et al. 2021, Mundhenk et al. 2021)
vars:
v1:
name: x1
sample_range:
- 0.0
- 3.0
sample_type:
- uni
- pos
v2:
name: x2
sample_range:
- 0.0
- 3.0
sample_type:
- uni
- pos
v3:
name: x3
sample_range:
- 0.0
- 3.0
sample_type:
- uni
- pos
v0:
name: y
meta:
dso_spec:
train:
all:
U:
- 0.0
- 3.0
- 100
test:
all:
U:
- 0.0
- 3.0
- 30
function_set: PKozaPlusSqrt
finite_fraction: 1.0
Poly-2:
raw: pow(x1,3)+pow(x1,2)+x1+sin(x1)+sin(pow(x2,2))
prepared: ((v1)**(3))+((v1)**(2))+v1+sin(v1)+sin(((v2)**(2)))
n_variables: 2
sources:
- Poli 2003 lineage (via DSO aggregation)
- formulas + ranges from dso-org/deep-symbolic-optimization benchmarks.csv (BSD-3;
Petersen et al. 2021, Mundhenk et al. 2021)
vars:
v1:
name: x1
sample_range:
- -1.0
- 1.0
sample_type:
- uni
- pos
v2:
name: x2
sample_range:
- -1.0
- 1.0
sample_type:
- uni
- pos
v0:
name: y
meta:
dso_spec:
train:
all:
U:
- -1
- 1
- 20
test: null
function_set: PKoza
finite_fraction: 1.0
Poly-3:
raw: cos(x2)/(sqrt(12*x1*x2+1.3+x1-0.05*pow(x2,2))+x1)
prepared: cos(v2)/(sqrt(12*v1*v2+1.3+v1-0.05*((v2)**(2)))+v1)
n_variables: 2
sources:
- Poli 2003 lineage (via DSO aggregation)
- formulas + ranges from dso-org/deep-symbolic-optimization benchmarks.csv (BSD-3;
Petersen et al. 2021, Mundhenk et al. 2021)
vars:
v1:
name: x1
sample_range:
- 0.0
- 3.0
sample_type:
- uni
- pos
v2:
name: x2
sample_range:
- 0.0
- 3.0
sample_type:
- uni
- pos
v0:
name: y
meta:
dso_spec:
train:
all:
U:
- 0.0
- 3.0
- 100
test:
all:
U:
- 0.0
- 3.0
- 30
function_set: PKozaPlusSqrt
finite_fraction: 1.0
Poly-4:
raw: sin(x4)/(sqrt(12*x1*x2+1.3-0.05*x3*x6*x10)*exp(x7))
prepared: sin(v4)/(sqrt(12*v1*v2+1.3-0.05*v3*v6*v10)*exp(v7))
n_variables: 10
sources:
- Poli 2003 lineage (via DSO aggregation)
- formulas + ranges from dso-org/deep-symbolic-optimization benchmarks.csv (BSD-3;
Petersen et al. 2021, Mundhenk et al. 2021)
vars:
v1:
name: x1
sample_range:
- 0.0
- 3.0
sample_type:
- uni
- pos
v2:
name: x2
sample_range:
- 0.0
- 3.0
sample_type:
- uni
- pos
v3:
name: x3
sample_range:
- 0.0
- 3.0
sample_type:
- uni
- pos
v4:
name: x4
sample_range:
- 0.0
- 3.0
sample_type:
- uni
- pos
v5:
name: x5
sample_range:
- 0.0
- 3.0
sample_type:
- uni
- pos
v6:
name: x6
sample_range:
- 0.0
- 3.0
sample_type:
- uni
- pos
v7:
name: x7
sample_range:
- 0.0
- 3.0
sample_type:
- uni
- pos
v8:
name: x8
sample_range:
- 0.0
- 3.0
sample_type:
- uni
- pos
v9:
name: x9
sample_range:
- 0.0
- 3.0
sample_type:
- uni
- pos
v10:
name: x10
sample_range:
- 0.0
- 3.0
sample_type:
- uni
- pos
v0:
name: y
meta:
dso_spec:
train:
all:
U:
- 0.0
- 3.0
- 330
test:
all:
U:
- 0.0
- 3.0
- 170
function_set: PKozaPlusSqrt
finite_fraction: 1.0
Poly-5:
raw: sin(pow(x1,3)-x1-pi/6)
prepared: sin(((v1)**(3))-v1-3.141592653589793/6)
n_variables: 1
sources:
- Poli 2003 lineage (via DSO aggregation)
- formulas + ranges from dso-org/deep-symbolic-optimization benchmarks.csv (BSD-3;
Petersen et al. 2021, Mundhenk et al. 2021)
vars:
v1:
name: x1
sample_range:
- -3.14
- 3.14
sample_type:
- uni
- pos
v0:
name: y
meta:
dso_spec:
train:
all:
U:
- -3.14
- 3.14
- 20
test: null
function_set: CPKoza
finite_fraction: 1.0