Note
Go to the end to download the full example code.
Quickstart: a first design#
Build a 30-run design for a thermal deposition study and read its quality evidence.
A materials-science lab is preparing a thermal deposition study on a multilayer coating. Reactor time is limited, so the goal is to pick a small set of runs that jointly cover the operating envelope as evenly as possible, ready to feed a first surrogate model.
Parameters#
temperature (100-500 degC, 10-degree steps): sweep the deposition window, from the lowest useful onset to the upper safe limit of the chamber.
pressure (0.5-5.0 bar, continuous): span sub-atmospheric to mildly pressurised conditions typical of chemical vapour deposition.
n_layers (2-10, integer): explore thin stacks up to the layer count the process can build in one campaign.
What to look at#
summary(): the “Optimised” and “Validation” counts confirm the design was built and that a hold-out set was reserved automatically.quality_report(): percentiles against the Monte Carlo baseline should be well above 90 for min_distance and near 0 for max_absolute_correlation; that pair is the primary evidence of a good space-filling design.plot('pairplot'): every 2D projection should look evenly populated, with no clustered corners or empty bands.
Mergen features used#
ParameterSpacewith three factor types (discrete, continuous, integer).Sampler.set_design()andSampler.run()at their defaults, so the example stays minimal and shows the shape of the workflow with no configuration to reason about.
Estimated runtime: a few seconds.
from mergen import ParameterSpace, Sampler
# 1. Define the parameter space
space = ParameterSpace({
'temperature': range(100, 500, 10), # discrete
'pressure': ('continuous', 0.5, 5.0), # continuous grid
'n_layers': ('integer', 2, 10), # integer grid
})
# 2. Configure and run the sampler
sampler = Sampler(space)
sampler.set_design()
result = sampler.run()
# 3. Inspect, visualise, export
result.summary()
result.quality_report()
result.plot('pairplot', save=True)
result.to_csv('design.csv')
════════════════════════════════════════════════════════════
MERGEN — Space-filling Design
════════════════════════════════════════════════════════════
Parameters : 3
Candidates : 36,000
n_samples : 30 (prescribed_in=0, focus_in=0, optimised_slots=30)
Total design : 30
Validation : 6
Criterion : umaxpro
Algorithm(s) : sa
────────────────────────────────────────────────────────────
[MERGEN] Optimising (criterion=umaxpro, algorithm=sa)...
[SA] Restart 1/5
[SA] Tuning temperature...
[SA] Start log(score)=35.870 T=5.2135e+20 iters=3000 swappable=30 hybrid=0.50
iter 500/3000 T=1.123e+20 best log(score)=34.675
iter 1000/3000 T=2.420e+19 best log(score)=34.675
iter 1500/3000 T=5.213e+18 best log(score)=34.675
iter 2000/3000 T=1.123e+18 best log(score)=34.675
iter 2500/3000 T=2.420e+17 best log(score)=34.675
iter 3000/3000 T=5.213e+16 best log(score)=34.675
[SA] Done log(score)=34.675 (accepted=2881/3000, rate=96.0%)
[SA] Restart 1: new best log(score)=34.675
[SA] Restart 2/5
[SA] Tuning temperature...
[SA] Start log(score)=47.598 T=1.5658e+21 iters=3000 swappable=30 hybrid=0.50
iter 500/3000 T=3.373e+20 best log(score)=34.139
iter 1000/3000 T=7.268e+19 best log(score)=34.139
iter 1500/3000 T=1.566e+19 best log(score)=34.139
iter 2000/3000 T=3.373e+18 best log(score)=34.139
iter 2500/3000 T=7.268e+17 best log(score)=34.139
iter 3000/3000 T=1.566e+17 best log(score)=34.139
[SA] Done log(score)=34.139 (accepted=2891/3000, rate=96.4%)
[SA] Restart 2: new best log(score)=34.139
[SA] Restart 3/5
[SA] Tuning temperature...
[SA] Start log(score)=49.134 T=3.5070e+20 iters=3000 swappable=30 hybrid=0.50
iter 500/3000 T=7.556e+19 best log(score)=36.093
iter 1000/3000 T=1.628e+19 best log(score)=35.979
iter 1500/3000 T=3.507e+18 best log(score)=35.979
iter 2000/3000 T=7.556e+17 best log(score)=35.144
iter 2500/3000 T=1.628e+17 best log(score)=35.144
iter 3000/3000 T=3.507e+16 best log(score)=35.144
[SA] Done log(score)=35.144 (accepted=2880/3000, rate=96.0%)
[SA] Restart 4/5
[SA] Tuning temperature...
[SA] Start log(score)=49.947 T=1.4001e+21 iters=3000 swappable=30 hybrid=0.50
iter 500/3000 T=3.017e+20 best log(score)=35.468
iter 1000/3000 T=6.499e+19 best log(score)=34.647
iter 1500/3000 T=1.400e+19 best log(score)=34.647
iter 2000/3000 T=3.017e+18 best log(score)=34.647
iter 2500/3000 T=6.499e+17 best log(score)=34.647
iter 3000/3000 T=1.400e+17 best log(score)=34.647
[SA] Done log(score)=34.647 (accepted=2871/3000, rate=95.7%)
[SA] Restart 5/5
[SA] Tuning temperature...
[SA] Start log(score)=50.211 T=1.1365e+21 iters=3000 swappable=30 hybrid=0.50
iter 500/3000 T=2.449e+20 best log(score)=34.994
iter 1000/3000 T=5.275e+19 best log(score)=34.631
iter 1500/3000 T=1.137e+19 best log(score)=34.631
iter 2000/3000 T=2.449e+18 best log(score)=34.631
iter 2500/3000 T=5.275e+17 best log(score)=34.631
iter 3000/3000 T=1.137e+17 best log(score)=34.631
[SA] Done log(score)=34.631 (accepted=2868/3000, rate=95.6%)
[MERGEN] sa done -- score=6.702e+14 (elapsed 29.3s)
────────────────────────────────────────────────────────────
MERGEN — Final Design
────────────────────────────────────────────────────────────
Prescribed (in) : 0
Prescribed (out) : 0
Focus (in) : 0
Focus (out) : 0
Optimised : 30
Total design : 30
Validation : 6
════════════════════════════════════════════════════════════
────────────────────────────────────────────────────
MERGEN Design Summary
────────────────────────────────────────────────────
Optimised : 30
Total design : 30
Validation : 6
────────────────────────────────────────────────────
Parameters : 3
Candidates : 36000
Criterion : umaxpro
Seed : 44
Algorithm : sa
────────────────────────────────────────────────────
[METRICS] Computing MC baseline (300 designs)...
[METRICS] MC baseline complete (300 designs).
════════════════════════════════════════════════════════════════════════
MERGEN Design Metrics (n=30, d=3)
════════════════════════════════════════════════════════════════════════
Metric Value Baseline Better when Rank
────────────────────────────────────────────────────────────────────────
Min distance 0.1355 0.0720 higher 97th pct
Minimax distance 0.4675 0.5416 lower 91th pct *
Max |correlation| 0.1968 0.2313 lower 63th pct *
2D projection CD2 0.0929 0.1174 lower 89th pct *
CV distances 0.3627 0.3709 lower 78th pct
Mean distance 0.6896 0.6937 higher 44th pct
────────────────────────────────────────────────────────────────────────
Criterion scores
────────────────────────────────────────────────────────────────────────
UMAXPRO 6.7021e+14 5.9234e+21 100th pct lower
────────────────────────────────────────────────────────────────────────
* = primarily optimised by 'umaxpro'
For other priorities, see: mergen.criteria.list_criteria()
────────────────────────────────────────────────────────────────────────
Baseline: 300 MC designs from feasible space | Rank = percentile among baseline designs
════════════════════════════════════════════════════════════════════════
Saved: outputs/pairplot_1.png
Saved: outputs/design.csv (36 rows)
Total running time of the script: (0 minutes 33.458 seconds)