Convert a database to a dataset

In this example, we will see how to convert a Database to a Dataset.

from __future__ import annotations

from gemseo import execute_algo
from gemseo.problems.optimization.rosenbrock import Rosenbrock

Let us solve the Rosenbrock optimization problem with the SLSQP algorithm and 10 iterations:

optimization_problem = Rosenbrock()
execute_algo(optimization_problem, algo_name="SLSQP", max_iter=10)

INFO - 16:24:39: Optimization problem:
INFO - 16:24:39:    minimize rosen(x) = sum( 100*(x[1:] - x[:-1]**2)**2 + (1 - x[:-1])**2 )
INFO - 16:24:39:    with respect to x
INFO - 16:24:39:    over the design space:
INFO - 16:24:39:       +------+-------------+-------+-------------+-------+
INFO - 16:24:39:       | Name | Lower bound | Value | Upper bound | Type  |
INFO - 16:24:39:       +------+-------------+-------+-------------+-------+
INFO - 16:24:39:       | x[0] |      -2     |   0   |      2      | float |
INFO - 16:24:39:       | x[1] |      -2     |   0   |      2      | float |
INFO - 16:24:39:       +------+-------------+-------+-------------+-------+
INFO - 16:24:39: Solving optimization problem with algorithm SLSQP:
INFO - 16:24:39:     10%|█         | 1/10 [00:00<00:00, 1045.44 it/sec, feas=True, obj=1]
INFO - 16:24:39:     20%|██        | 2/10 [00:00<00:00, 1559.80 it/sec, feas=True, obj=1.6e+3]
INFO - 16:24:39:     30%|███       | 3/10 [00:00<00:00, 1804.26 it/sec, feas=True, obj=0.8]
INFO - 16:24:39:     40%|████      | 4/10 [00:00<00:00, 1827.78 it/sec, feas=True, obj=962]
INFO - 16:24:39:     50%|█████     | 5/10 [00:00<00:00, 1814.30 it/sec, feas=True, obj=0.681]
INFO - 16:24:39:     60%|██████    | 6/10 [00:00<00:00, 1840.01 it/sec, feas=True, obj=8.3]
INFO - 16:24:39:     70%|███████   | 7/10 [00:00<00:00, 1885.69 it/sec, feas=True, obj=6.17]
INFO - 16:24:39:     80%|████████  | 8/10 [00:00<00:00, 1872.77 it/sec, feas=True, obj=0.398]
INFO - 16:24:39:     90%|█████████ | 9/10 [00:00<00:00, 1863.95 it/sec, feas=True, obj=0.221]
INFO - 16:24:39:    100%|██████████| 10/10 [00:00<00:00, 1856.71 it/sec, feas=True, obj=0.193]
INFO - 16:24:39: Optimization result:
INFO - 16:24:39:    Optimizer info:
INFO - 16:24:39:       Status: None
INFO - 16:24:39:       Message: Maximum number of iterations reached. GEMSEO stopped the driver.
INFO - 16:24:39:    Solution:
INFO - 16:24:39:       Objective: 0.19349100516217954
INFO - 16:24:39:       Design space:
INFO - 16:24:39:          +------+-------------+--------------------+-------------+-------+
INFO - 16:24:39:          | Name | Lower bound |       Value        | Upper bound | Type  |
INFO - 16:24:39:          +------+-------------+--------------------+-------------+-------+
INFO - 16:24:39:          | x[0] |      -2     | 0.5706658475298814 |      2      | float |
INFO - 16:24:39:          | x[1] |      -2     | 0.316087053903805  |      2      | float |
INFO - 16:24:39:          +------+-------------+--------------------+-------------+-------+

Optimization result:

• Design variables: [0.57066585 0.31608705]
• Objective function: 0.19349100516217954
• Feasible solution: True

Then, the Database attached to this OptimizationProblem can be converted to an OptimizationDataset using its method to_dataset():

dataset = optimization_problem.to_dataset()
dataset

GROUP	designs		functions
VARIABLE	x		rosen
COMPONENT	0	1	0
1	0.000000	0.000000	1.000000
2	2.000000	0.000000	1601.000000
3	0.200000	0.000000	0.800000
4	2.000000	0.899368	962.392108
5	0.380000	0.089937	0.681024
6	1.309049	2.000000	8.297426
7	0.682846	0.712569	6.166440
8	0.449476	0.232774	0.397606
9	0.551433	0.290027	0.220957
10	0.570666	0.316087	0.193491

The design variables and output variables are in separate groups. You can also use an IODataset instead of an OptimizationDataset:

dataset = optimization_problem.to_dataset(opt_naming=False)
dataset

GROUP	inputs		outputs
VARIABLE	x		rosen
COMPONENT	0	1	0
0	0.000000	0.000000	1.000000
1	2.000000	0.000000	1601.000000
2	0.200000	0.000000	0.800000
3	2.000000	0.899368	962.392108
4	0.380000	0.089937	0.681024
5	1.309049	2.000000	8.297426
6	0.682846	0.712569	6.166440
7	0.449476	0.232774	0.397606
8	0.551433	0.290027	0.220957
9	0.570666	0.316087	0.193491

or simply do not separate the variables

dataset = optimization_problem.to_dataset(categorize=False)
dataset

GROUP	parameters
VARIABLE	x		rosen
COMPONENT	0	1	0
0	0.000000	0.000000	1.000000
1	2.000000	0.000000	1601.000000
2	0.200000	0.000000	0.800000
3	2.000000	0.899368	962.392108
4	0.380000	0.089937	0.681024
5	1.309049	2.000000	8.297426
6	0.682846	0.712569	6.166440
7	0.449476	0.232774	0.397606
8	0.551433	0.290027	0.220957
9	0.570666	0.316087	0.193491

Note

Only design variables and functions (objective function, constraints) are stored in the database. If you want to store state variables, you must add them as observables before the problem is executed. Use the add_observable() method.