Reporting and Analysis

Extracting, formatting, and analyzing optimization results.

After an optimization run, spotoptim provides utilities for printing the best solution, building summary tables, and computing variable importance and sensitivity. All functions accept a SpotOptim instance and work on the data stored during optimize().

Printing the Best Solution

print_best displays the best parameter vector, objective value, and total evaluation count in a human-readable format. Factor variables are mapped back to their original string labels.

import numpy as np
from spotoptim import SpotOptim
from spotoptim.function import sphere
from spotoptim.reporting.results import print_best

opt = SpotOptim(
    fun=sphere,
    bounds=[(-5, 5), (-5, 5)],
    max_iter=25,
    n_initial=10,
    seed=0,
)
result = opt.optimize()

print_best(opt, result)

Best Solution Found:
--------------------------------------------------
  x0: -0.0002
  x1: 0.0007
  Objective Value: 0.0000
  Total Evaluations: 25

Pass transformations (a list of callables, one per dimension) to display parameters on a different scale — for example, 10**x for log-transformed learning rates.

Results Table

get_results_table returns a formatted table string with one row per variable showing its name, type, bounds, default value, and tuned (best) value. Set show_importance=True to append importance scores and star ratings.

import numpy as np
from spotoptim import SpotOptim
from spotoptim.function import sphere
from spotoptim.reporting.results import get_results_table

opt = SpotOptim(
    fun=sphere,
    bounds=[(-5, 5), (-5, 5)],
    max_iter=25,
    n_initial=10,
    seed=0,
    var_name=["x1", "x2"],
)
result = opt.optimize()

table = get_results_table(opt)
print(table)
|   name |   type |   default |   lower |   upper |   tuned |   transform |
|--------|--------|-----------|---------|---------|---------|-------------|
|     x1 |  float |         0 |      -5 |       5 | -0.0002 |           - |
|     x2 |  float |         0 |      -5 |       5 |  0.0007 |           - |

The tablefmt argument accepts any format supported by the tabulate library (e.g. "github", "pipe", "grid", "latex").

Design Table

get_design_table summarises the search space before optimization results are known: variable names, types, bounds, default (midpoint) values, and transformations.

import numpy as np
from spotoptim import SpotOptim
from spotoptim.function import sphere
from spotoptim.reporting.results import get_design_table

opt = SpotOptim(
    fun=sphere,
    bounds=[(-5, 5), (-5, 5)],
    max_iter=25,
    n_initial=10,
    seed=0,
    var_name=["x1", "x2"],
)
result = opt.optimize()

design = get_design_table(opt)
print(design)
|   name |   type |   lower |   upper |   default |   transform |
|--------|--------|---------|---------|-----------|-------------|
|     x1 |  float |      -5 |       5 |         0 |           - |
|     x2 |  float |      -5 |       5 |         0 |           - |

Variable Importance

get_importance computes a correlation-based importance score for each variable. Scores are normalised to a 0–100 scale: a variable that correlates strongly with the objective receives a high score, while a variable with no correlation scores near zero.

import numpy as np
from spotoptim import SpotOptim
from spotoptim.function import sphere
from spotoptim.reporting.analysis import get_importance

opt = SpotOptim(
    fun=sphere,
    bounds=[(-5, 5), (-5, 5)],
    max_iter=25,
    n_initial=10,
    seed=0,
    var_name=["x1", "x2"],
)
result = opt.optimize()

importance = get_importance(opt)
for name, score in zip(["x1", "x2"], importance):
    print(f"{name}: {score:.1f}%")
x1: 41.0%
x2: 59.0%

The sphere function treats both dimensions symmetrically, so their importance scores will be roughly equal.

Sensitivity Analysis

sensitivity_spearman prints Spearman rank correlations between each parameter and the objective value, together with p-values and significance stars:

  • *** — \(p < 0.001\) (highly significant)
  • ** — \(p < 0.01\)
  • * — \(p < 0.05\)
import numpy as np
from spotoptim import SpotOptim
from spotoptim.function import sphere
from spotoptim.reporting.analysis import sensitivity_spearman

opt = SpotOptim(
    fun=sphere,
    bounds=[(-5, 5), (-5, 5)],
    max_iter=25,
    n_initial=10,
    seed=0,
    var_name=["x1", "x2"],
)
result = opt.optimize()

sensitivity_spearman(opt)
print("Done")

Sensitivity Analysis (Spearman Correlation):
--------------------------------------------------
  x1                  : +0.026 (p=0.901)
  x2                  : +0.121 (p=0.565)
Done

Next Steps