Self-Organizing Map#

In this example, we illustrate the use of the SOM plot on the Sobieski's SSBJ problem.

from __future__ import annotations

from gemseo import configure_logger
from gemseo import create_discipline
from gemseo import create_scenario
from gemseo.problems.mdo.sobieski.core.design_space import SobieskiDesignSpace

Import#

The first step is to import some high-level functions and a method to get the design space.

configure_logger()
<RootLogger root (INFO)>

Description#

The SOM post-processing performs a Self Organizing Map clustering on the optimization history. A SOM is a 2D representation of a design of experiments which requires dimensionality reduction since it may be in a very high dimension.

A SOM is built by using an unsupervised artificial neural network [KSH01]. A map of size n_x.n_y is generated, where n_x is the number of neurons in the \(x\) direction and n_y is the number of neurons in the \(y\) direction. The design space (whatever the dimension) is reduced to a 2D representation based on n_x.n_y neurons. Samples are clustered to a neuron when their design variables are close in terms of their L2 norm. A neuron is always located at the same place on a map. Each neuron is colored according to the average value for a given criterion. This helps to qualitatively analyze whether parts of the design space are good according to some criteria and not for others, and where compromises should be made. A white neuron has no sample associated with it: not enough evaluations were provided to train the SOM.

SOM's provide a qualitative view of the objective function, the constraints, and of their relative behaviors.

Create disciplines#

At this point, we instantiate the disciplines of Sobieski's SSBJ problem: Propulsion, Aerodynamics, Structure and Mission

disciplines = create_discipline([
    "SobieskiPropulsion",
    "SobieskiAerodynamics",
    "SobieskiStructure",
    "SobieskiMission",
])

Create design space#

We also create the SobieskiDesignSpace.

design_space = SobieskiDesignSpace()

Create and execute scenario#

The next step is to build an MDO scenario in order to maximize the range, encoded 'y_4', with respect to the design parameters, while satisfying the inequality constraints 'g_1', 'g_2' and 'g_3'. We can use the MDF formulation, the Monte Carlo DOE algorithm and 30 samples.

scenario = create_scenario(
    disciplines,
    "y_4",
    design_space,
    formulation_name="MDF",
    maximize_objective=True,
    scenario_type="DOE",
)
for constraint in ["g_1", "g_2", "g_3"]:
    scenario.add_constraint(constraint, constraint_type="ineq")
scenario.execute(algo_name="OT_MONTE_CARLO", n_samples=30)
WARNING - 08:39:45: Unsupported feature 'minItems' in JSONGrammar 'SobieskiMission_discipline_output' for property 'y_4' in conversion to SimpleGrammar.
WARNING - 08:39:45: Unsupported feature 'maxItems' in JSONGrammar 'SobieskiMission_discipline_output' for property 'y_4' in conversion to SimpleGrammar.
   INFO - 08:39:45:
   INFO - 08:39:45: *** Start DOEScenario execution ***
   INFO - 08:39:45: DOEScenario
   INFO - 08:39:45:    Disciplines: SobieskiAerodynamics SobieskiMission SobieskiPropulsion SobieskiStructure
   INFO - 08:39:45:    MDO formulation: MDF
   INFO - 08:39:45: Optimization problem:
   INFO - 08:39:45:    minimize -y_4(x_shared, x_1, x_2, x_3)
   INFO - 08:39:45:    with respect to x_1, x_2, x_3, x_shared
   INFO - 08:39:45:    subject to constraints:
   INFO - 08:39:45:       g_1(x_shared, x_1, x_2, x_3) <= 0
   INFO - 08:39:45:       g_2(x_shared, x_1, x_2, x_3) <= 0
   INFO - 08:39:45:       g_3(x_shared, x_1, x_2, x_3) <= 0
   INFO - 08:39:45:    over the design space:
   INFO - 08:39:45:       +-------------+-------------+-------+-------------+-------+
   INFO - 08:39:45:       | Name        | Lower bound | Value | Upper bound | Type  |
   INFO - 08:39:45:       +-------------+-------------+-------+-------------+-------+
   INFO - 08:39:45:       | x_shared[0] |     0.01    |  0.05 |     0.09    | float |
   INFO - 08:39:45:       | x_shared[1] |    30000    | 45000 |    60000    | float |
   INFO - 08:39:45:       | x_shared[2] |     1.4     |  1.6  |     1.8     | float |
   INFO - 08:39:45:       | x_shared[3] |     2.5     |  5.5  |     8.5     | float |
   INFO - 08:39:45:       | x_shared[4] |      40     |   55  |      70     | float |
   INFO - 08:39:45:       | x_shared[5] |     500     |  1000 |     1500    | float |
   INFO - 08:39:45:       | x_1[0]      |     0.1     |  0.25 |     0.4     | float |
   INFO - 08:39:45:       | x_1[1]      |     0.75    |   1   |     1.25    | float |
   INFO - 08:39:45:       | x_2         |     0.75    |   1   |     1.25    | float |
   INFO - 08:39:45:       | x_3         |     0.1     |  0.5  |      1      | float |
   INFO - 08:39:45:       +-------------+-------------+-------+-------------+-------+
   INFO - 08:39:45: Solving optimization problem with algorithm OT_MONTE_CARLO:
   INFO - 08:39:45:      3%|▎         | 1/30 [00:00<00:02, 13.84 it/sec, obj=-166]
   INFO - 08:39:45:      7%|▋         | 2/30 [00:00<00:01, 18.18 it/sec, obj=-484]
   INFO - 08:39:45:     10%|█         | 3/30 [00:00<00:01, 20.30 it/sec, obj=-481]
   INFO - 08:39:45:     13%|█▎        | 4/30 [00:00<00:01, 21.63 it/sec, obj=-384]
   INFO - 08:39:45:     17%|█▋        | 5/30 [00:00<00:01, 22.49 it/sec, obj=-1.14e+3]
   INFO - 08:39:45:     20%|██        | 6/30 [00:00<00:01, 23.11 it/sec, obj=-290]
   INFO - 08:39:45:     23%|██▎       | 7/30 [00:00<00:00, 23.21 it/sec, obj=-630]
   INFO - 08:39:45:     27%|██▋       | 8/30 [00:00<00:00, 23.00 it/sec, obj=-346]
   INFO - 08:39:45:     30%|███       | 9/30 [00:00<00:00, 23.08 it/sec, obj=-626]
   INFO - 08:39:45:     33%|███▎      | 10/30 [00:00<00:00, 23.16 it/sec, obj=-621]
   INFO - 08:39:45:     37%|███▋      | 11/30 [00:00<00:00, 23.22 it/sec, obj=-280]
   INFO - 08:39:45:     40%|████      | 12/30 [00:00<00:00, 22.67 it/sec, obj=-288]
   INFO - 08:39:45:     43%|████▎     | 13/30 [00:00<00:00, 22.22 it/sec, obj=-257]
   INFO - 08:39:45:     47%|████▋     | 14/30 [00:00<00:00, 21.85 it/sec, obj=-367]
   INFO - 08:39:45:     50%|█████     | 15/30 [00:00<00:00, 21.69 it/sec, obj=-1.08e+3]
   INFO - 08:39:45:     53%|█████▎    | 16/30 [00:00<00:00, 21.81 it/sec, obj=-344]
   INFO - 08:39:46:     57%|█████▋    | 17/30 [00:00<00:00, 21.63 it/sec, obj=-368]
   INFO - 08:39:46:     60%|██████    | 18/30 [00:00<00:00, 21.50 it/sec, obj=-253]
   INFO - 08:39:46:     63%|██████▎   | 19/30 [00:00<00:00, 21.29 it/sec, obj=-129]
   INFO - 08:39:46:     67%|██████▋   | 20/30 [00:00<00:00, 21.21 it/sec, obj=-1.07e+3]
   INFO - 08:39:46:     70%|███████   | 21/30 [00:00<00:00, 21.43 it/sec, obj=-341]
   INFO - 08:39:46:     73%|███████▎  | 22/30 [00:01<00:00, 21.53 it/sec, obj=-1e+3]
   INFO - 08:39:46:     77%|███████▋  | 23/30 [00:01<00:00, 21.18 it/sec, obj=-586]
   INFO - 08:39:46:     80%|████████  | 24/30 [00:01<00:00, 21.28 it/sec, obj=-483]
   INFO - 08:39:46:     83%|████████▎ | 25/30 [00:01<00:00, 21.46 it/sec, obj=-392]
   INFO - 08:39:46:     87%|████████▋ | 26/30 [00:01<00:00, 21.62 it/sec, obj=-406]
   INFO - 08:39:46:     90%|█████████ | 27/30 [00:01<00:00, 21.45 it/sec, obj=-207]
   INFO - 08:39:46:     93%|█████████▎| 28/30 [00:01<00:00, 21.61 it/sec, obj=-702]
   INFO - 08:39:46:     97%|█████████▋| 29/30 [00:01<00:00, 21.83 it/sec, obj=-423]
   INFO - 08:39:46:    100%|██████████| 30/30 [00:01<00:00, 21.90 it/sec, obj=-664]
   INFO - 08:39:46: Optimization result:
   INFO - 08:39:46:    Optimizer info:
   INFO - 08:39:46:       Status: None
   INFO - 08:39:46:       Message: None
   INFO - 08:39:46:       Number of calls to the objective function by the optimizer: 30
   INFO - 08:39:46:    Solution:
   INFO - 08:39:46:       The solution is feasible.
   INFO - 08:39:46:       Objective: -367.45728393799953
   INFO - 08:39:46:       Standardized constraints:
   INFO - 08:39:46:          g_1 = [-0.02478574 -0.00310924 -0.00855146 -0.01702654 -0.02484732 -0.04764585
   INFO - 08:39:46:  -0.19235415]
   INFO - 08:39:46:          g_2 = -0.09000000000000008
   INFO - 08:39:46:          g_3 = [-0.98722984 -0.01277016 -0.60760341 -0.0557087 ]
   INFO - 08:39:46:       Design space:
   INFO - 08:39:46:          +-------------+-------------+---------------------+-------------+-------+
   INFO - 08:39:46:          | Name        | Lower bound |        Value        | Upper bound | Type  |
   INFO - 08:39:46:          +-------------+-------------+---------------------+-------------+-------+
   INFO - 08:39:46:          | x_shared[0] |     0.01    | 0.01230934749207792 |     0.09    | float |
   INFO - 08:39:46:          | x_shared[1] |    30000    |  43456.87364611478  |    60000    | float |
   INFO - 08:39:46:          | x_shared[2] |     1.4     |  1.731884935123487  |     1.8     | float |
   INFO - 08:39:46:          | x_shared[3] |     2.5     |  3.894765253193514  |     8.5     | float |
   INFO - 08:39:46:          | x_shared[4] |      40     |  57.92631048228255  |      70     | float |
   INFO - 08:39:46:          | x_shared[5] |     500     |  520.4048463450415  |     1500    | float |
   INFO - 08:39:46:          | x_1[0]      |     0.1     |  0.3994784918586811 |     0.4     | float |
   INFO - 08:39:46:          | x_1[1]      |     0.75    |  0.9500312867674923 |     1.25    | float |
   INFO - 08:39:46:          | x_2         |     0.75    |  1.205851870260564  |     1.25    | float |
   INFO - 08:39:46:          | x_3         |     0.1     |  0.2108042391973412 |      1      | float |
   INFO - 08:39:46:          +-------------+-------------+---------------------+-------------+-------+
   INFO - 08:39:46: *** End DOEScenario execution (time: 0:00:01.380745) ***

Post-process scenario#

Lastly, we post-process the scenario by means of the SOM plot which performs a self organizing map clustering on optimization history.

Tip

Each post-processing method requires different inputs and offers a variety of customization options. Use the high-level function get_post_processing_options_schema() to print a table with the options for any post-processing algorithm. Or refer to our dedicated page: Post-processing algorithms.

scenario.post_process(post_name="SOM", save=False, show=True)
Self Organizing Maps of the design space, -y_4, g_1[0], g_1[1], g_1[2], g_1[3], g_1[4], g_1[5], g_1[6], g_2, g_3[0], g_3[1], g_3[2], g_3[3]
    INFO - 08:39:46: Building Self Organizing Map from optimization history:
    INFO - 08:39:46:     Number of neurons in x direction = 4
    INFO - 08:39:46:     Number of neurons in y direction = 4

<gemseo.post.som.SOM object at 0x7f24dcb04310>

The following figure illustrates another SOM on the Sobieski use case. The optimization method is a (costly) derivative free algorithm (NLOPT_COBYLA), indeed all the relevant information for the optimization is obtained at the cost of numerous evaluations of the functions. For more details, please read the paper by [KJO+06] on wing MDO post-processing using SOM.

../../../_images/MDOScenario_SOM_v100.png

SOM example on the Sobieski problem.#

A DOE may also be a good way to produce SOM maps. The following figure shows an example with 10000 points on the same test case. This produces more relevant SOM plots.

../../../_images/som_fine.png

SOM example on the Sobieski problem with a 10 000 samples DOE.#

Total running time of the script: (0 minutes 2.661 seconds)

Gallery generated by Sphinx-Gallery