v0.2.3 More examples (engineering problems, ML models hyper-parameters tuning)

Author: CameleoGrey

README.md

@@ -44,14 +44,14 @@ maturin develop --release
 -  maturin will build the Rust part, get all Python dependencies (for solver itself, not examples) and install greyjack to your venv
 
 # RoadMap
-- CMA, probably its modern variants, adaptations for tasks with integer and categorical variables (often appears in articles as sota approach)
 - Add more examples: Job-Shop, Pickup&Delivery, some continuous and MIP tasks, scheduling, etc
+- Types, arguments validation
+- Write docs
+- Tests, tests, tests...
 - Composite termination criterion (for example: solving limit minutes N AND score not improving M seconds)
 - Multi-level score
 - Custom moves support
 - Try to impove incremental (pseudo-incremental) score calculation mechanism (caching, no clonning, etc)
-- Write docs
 - Website
 - Useful text materials, guides, presentations
-- Tests, tests, tests...
-- Score explainer / interpreter
+- Score explainer / interpreter for OOP API

examples/pure_math/engineering/__init__.py

@@ -0,0 +1,56 @@
+from copy import deepcopy
+from pathlib import Path
+import random
+import os
+import sys
+
+# To launch normally from console
+script_dir_path = Path(os.path.dirname(os.path.realpath(__file__)))
+project_dir_id = script_dir_path.parts.index("greyjack-solver-python")
+project_dir_path = Path(*script_dir_path.parts[:project_dir_id+1])
+sys.path.append(str(project_dir_path))
+
+
+from copy import deepcopy
+
+from greyjack.SolverPureMath import SolverPureMath
+from greyjack.agents import *
+from greyjack.agents.termination_strategies import *
+from greyjack.agents.base.ParallelizationBackend import ParallelizationBackend
+from greyjack.agents.base.LoggingLevel import LoggingLevel
+
+from greyjack.pure_math.MathModel import MathModel
+from greyjack.pure_math.Objective import Objective
+from greyjack.pure_math.Constraint import Constraint
+from greyjack.pure_math.variables.FloatVar import FloatVar
+from greyjack.pure_math.variables.IntegerVar import IntegerVar
+from greyjack.pure_math.variables.BinaryVar import BinaryVar
+
+"""
+https://www.sci-hub.ru/10.1016/j.cie.2021.107250
+"""
+
+if __name__ == "__main__":
+    m = math_model = MathModel()
+
+    for i in range( 1, 6 ): 
+        m.variables[f"x{i}"] = FloatVar( 0.01, 100 )
+    m.objectives["f"] = Objective("minimize", lambda v, u: 0.06224*sum([x_i for x_i in v.values()]))
+    m.constraints["g"] = Constraint( lambda v, u: 60/v["x1"]**3 + 27/v["x2"]**3 + 19/v["x3"]**3 + 7/v["x4"]**3 + 1/v["x5"]**3 - 1, "<=")
+
+    termination_strategy = ScoreNoImprovement(time_seconds_limit=5)
+    agent = GeneticAlgorithm(population_size=256, crossover_probability=0.8, p_best_rate=0.5,
+                             tabu_entity_rate=0.2, mutation_rate_multiplier=1.0, move_probas=[0.5, 0.5, 0.0, 0.0, 0.0, 0.0],
+                             migration_rate=0.00001, migration_frequency=10, termination_strategy=termination_strategy)
+    """agent = LateAcceptance(late_acceptance_size=200, tabu_entity_rate=0.2, 
+                           mutation_rate_multiplier=1.0, move_probas=[0.5, 0.5, 0.0, 0.0, 0.0, 0.0], 
+                           compare_to_global_frequency=10000, termination_strategy=termination_strategy)"""
+    
+    solver = SolverPureMath(math_model, agent,
+                            ParallelizationBackend.Multiprocessing, LoggingLevel.FreshOnly,
+                            n_jobs=10, score_precision=[4, 2])
+    solution = solver.solve()
+    math_model.explain_solution( solution )
+
+    print( solution )
+    print( "done" )

examples/pure_math/engineering/cantilever_beam_design.py

@@ -0,0 +1,94 @@
+from copy import deepcopy
+from pathlib import Path
+import random
+import os
+import sys
+
+# To launch normally from console
+script_dir_path = Path(os.path.dirname(os.path.realpath(__file__)))
+project_dir_id = script_dir_path.parts.index("greyjack-solver-python")
+project_dir_path = Path(*script_dir_path.parts[:project_dir_id+1])
+sys.path.append(str(project_dir_path))
+
+from greyjack.SolverPureMath import SolverPureMath
+from greyjack.agents import *
+from greyjack.agents.termination_strategies import *
+from greyjack.agents.base.ParallelizationBackend import ParallelizationBackend
+from greyjack.agents.base.LoggingLevel import LoggingLevel
+from greyjack.pure_math.MathModel import MathModel
+from greyjack.pure_math.Objective import Objective
+from greyjack.pure_math.Constraint import Constraint
+from greyjack.pure_math.variables.FloatVar import FloatVar
+from greyjack.pure_math.variables.IntegerVar import IntegerVar
+from greyjack.pure_math.variables.BinaryVar import BinaryVar
+
+from copy import deepcopy
+from math import pi
+
+"""
+Strange problem. Different papers "lose" some coefficients, have different formulations.
+
+links:
+https://www.researchgate.net/publication/352203707_Enhanced_grasshopper_optimization_algorithm_using_elite_opposition-based_learning_for_solving_real-world_engineering_problems
+https://www.sci-hub.ru/10.1016/j.cie.2021.107250
+"""
+
+if __name__ == "__main__":
+    m = math_model = MathModel()
+
+    m.variables["t"] = FloatVar(1.5, 3)
+    m.variables["r_o"] = IntegerVar(90, 110)
+    m.variables["r_i"] = IntegerVar(60, 80)
+    m.variables["F"] = IntegerVar(600, 1000)
+    m.variables["Z"] = IntegerVar(2, 9)
+
+    m.utility["n"] = 250 #rpm
+    m.utility["M_f"] = 3 #Nm
+    m.utility["M_s"] = 40 #Nm
+    m.utility["I_z"] = 55 #kg/m^2
+    m.utility["dr"] = 20 #mm
+    m.utility["t_min"] = 1.5 #mm
+    m.utility["t_max"] = 3 #mm
+    m.utility["l_max"] = 30 #mm
+    m.utility["Z_max"] = 9
+    m.utility["v_sr_max"] = 10 #m/s
+    m.utility["nu"] = 0.5
+    m.utility["delta"] = 0.5
+    m.utility["p_max"] = 1 #MPa
+    m.utility["T_max"] = 15 #s
+    m.utility["F_max"] = 1000 #N
+    m.utility["r_i_min"] = 60 #mm
+    m.utility["r_i_max"] = 80 #mm
+    m.utility["r_o_min"] = 90 #mm
+    m.utility["r_o_max"] = 110 #mm
+    m.utility["ro"] = 7.8e-6
+    m.utility["s"] = 1.5
+    m.utility["M_h"] = lambda v, u: ((2 / 3) * u["nu"] * v["F"] * v["Z"] * (v["r_o"]**3 - v["r_i"]**3)) / (v["r_o"]**2 - v["r_i"]**2)
+    m.utility["p_rz"] = lambda v, u: v["F"] / (pi * (v["r_o"]**2 - v["r_i"]**2))
+    m.utility["v_sr"] = lambda v, u: (2 * pi * ((v["r_o"]**3 - v["r_i"]**3)) / (90 * (v["r_o"]**2 - v["r_i"]**2)))
+    m.utility["T"] = lambda v, u: (u["I_z"] * pi * u["n"]) / (30 * (u["M_h"](v, u) + u["M_f"]))
+
+    m.objectives["f"] = Objective("minimize", lambda v, u: pi*v["t"]*u["ro"]*(v["r_o"]**2 - v["r_i"]**2)*(v["Z"] + 1))
+
+    m.constraints["g_1"] = Constraint( lambda v, u: v["r_o"] - v["r_i"] - u["dr"], ">=")
+    m.constraints["g_2"] = Constraint( lambda v, u: u["l_max"] - (v["Z"]+1)*(v["t"]+u["delta"]), ">=")
+    m.constraints["g_3"] = Constraint( lambda v, u: u["p_max"] + u["p_rz"](v, u), ">=")
+    m.constraints["g_4"] = Constraint( lambda v, u: u["p_max"]*u["v_sr_max"] - u["p_rz"](v, u)*u["v_sr"](v, u), ">=")
+    m.constraints["g_5"] = Constraint( lambda v, u: u["v_sr_max"] - u["v_sr"](v, u), ">=")
+    m.constraints["g_6"] = Constraint( lambda v, u: u["T_max"] - u["T"](v, u), ">=")
+    m.constraints["g_7"] = Constraint( lambda v, u: u["M_h"](v, u) - u["s"]*u["M_s"], ">=")
+    m.constraints["g_8"] = Constraint( lambda v, u: u["T"](v, u), ">=")
+
+    termination_strategy = ScoreNoImprovement(time_seconds_limit=5)
+    agent = GeneticAlgorithm(population_size=256, crossover_probability=0.8, p_best_rate=0.5,
+                             tabu_entity_rate=0.2, mutation_rate_multiplier=1.0, move_probas=[0.5, 0.5, 0.0, 0.0, 0.0, 0.0],
+                             migration_rate=0.00001, migration_frequency=10, termination_strategy=termination_strategy)
+    
+    solver = SolverPureMath(math_model, agent,
+                            ParallelizationBackend.Multiprocessing, LoggingLevel.FreshOnly,
+                            n_jobs=10, score_precision=[4, 2])
+    solution = solver.solve()
+    math_model.explain_solution( solution )
+
+    print( solution )
+    print( "done" )

examples/pure_math/engineering/multiple_disk_clutch_brake_design.py

@@ -0,0 +1,67 @@
+from copy import deepcopy
+from pathlib import Path
+import random
+import os
+import sys
+
+# To launch normally from console
+script_dir_path = Path(os.path.dirname(os.path.realpath(__file__)))
+project_dir_id = script_dir_path.parts.index("greyjack-solver-python")
+project_dir_path = Path(*script_dir_path.parts[:project_dir_id+1])
+sys.path.append(str(project_dir_path))
+
+from greyjack.SolverPureMath import SolverPureMath
+from greyjack.agents import *
+from greyjack.agents.termination_strategies import *
+from greyjack.agents.base.ParallelizationBackend import ParallelizationBackend
+from greyjack.agents.base.LoggingLevel import LoggingLevel
+
+from greyjack.pure_math.MathModel import MathModel
+from greyjack.pure_math.Objective import Objective
+from greyjack.pure_math.Constraint import Constraint
+from greyjack.pure_math.variables.FloatVar import FloatVar
+from greyjack.pure_math.variables.IntegerVar import IntegerVar
+from greyjack.pure_math.variables.BinaryVar import BinaryVar
+
+from copy import deepcopy
+from math import pi
+
+"""
+https://www.sci-hub.ru/10.1016/j.cie.2021.107250
+https://ceur-ws.org/Vol-2255/paper2.pdf
+"""
+
+if __name__ == "__main__":
+    m = math_model = MathModel()
+
+    m.variables["x1"] = FloatVar( 0, 99 )
+    m.variables["x2"] = FloatVar( 0, 99 )
+    m.variables["x3"] = FloatVar( 10, 200 )
+    m.variables["x4"] = FloatVar( 10, 200 )
+
+    # optimum objective in https://ceur-ws.org/Vol-2255/paper2.pdf is 0.0126652327883
+    m.objectives["f"] = Objective("minimize",
+                                lambda v, u: 0.6224*v["x1"]*v["x3"]*v["x4"] + 1.7781*v["x2"]*v["x3"]**2 +
+                                            3.1661*v["x1"]**2*v["x4"] + 19.84*v["x1"]**2*v["x3"] )
+
+    m.constraints["g1"] = Constraint( lambda v, u: -v["x1"] + 0.0193*v["x3"], "<=")
+    m.constraints["g2"] = Constraint( lambda v, u: -v["x2"] + 0.00954*v["x3"], "<=")
+    m.constraints["g3"] = Constraint( lambda v, u: -pi*v["x3"]**2*v["x4"] - (4/3)*pi*v["x3"]**3 + 1296000, "<=")
+    m.constraints["g4"] = Constraint( lambda v, u: v["x4"] - 240, "<=")
+
+    termination_strategy = ScoreNoImprovement(time_seconds_limit=5)
+    """agent = GeneticAlgorithm(population_size=256, crossover_probability=0.8, p_best_rate=0.5,
+                             tabu_entity_rate=0.2, mutation_rate_multiplier=1.0, move_probas=[0.5, 0.5, 0.0, 0.0, 0.0, 0.0],
+                             migration_rate=0.00001, migration_frequency=10, termination_strategy=termination_strategy)"""
+    agent = LateAcceptance(late_acceptance_size=200, tabu_entity_rate=0.2, 
+                           mutation_rate_multiplier=1.0, move_probas=[0.5, 0.5, 0.0, 0.0, 0.0, 0.0], 
+                           compare_to_global_frequency=10000, termination_strategy=termination_strategy)
+    
+    solver = SolverPureMath(math_model, agent,
+                            ParallelizationBackend.Multiprocessing, LoggingLevel.FreshOnly,
+                            n_jobs=10, score_precision=[4, 2])
+    solution = solver.solve()
+    math_model.explain_solution( solution )
+
+    print( solution )
+    print( "done" )

examples/pure_math/engineering/pressure_vessel_design.py

@@ -0,0 +1,81 @@
+from copy import deepcopy
+from pathlib import Path
+import random
+import os
+import sys
+
+# To launch normally from console
+script_dir_path = Path(os.path.dirname(os.path.realpath(__file__)))
+project_dir_id = script_dir_path.parts.index("greyjack-solver-python")
+project_dir_path = Path(*script_dir_path.parts[:project_dir_id+1])
+sys.path.append(str(project_dir_path))
+
+
+from copy import deepcopy
+
+from greyjack.SolverPureMath import SolverPureMath
+from greyjack.agents import *
+from greyjack.agents.termination_strategies import *
+from greyjack.agents.base.ParallelizationBackend import ParallelizationBackend
+from greyjack.agents.base.LoggingLevel import LoggingLevel
+
+from greyjack.pure_math.MathModel import MathModel
+from greyjack.pure_math.Objective import Objective
+from greyjack.pure_math.Constraint import Constraint
+from greyjack.pure_math.variables.FloatVar import FloatVar
+from greyjack.pure_math.variables.IntegerVar import IntegerVar
+from greyjack.pure_math.variables.BinaryVar import BinaryVar
+
+"""
+https://www.sci-hub.ru/10.1016/j.cie.2021.107250
+"""
+
+if __name__ == "__main__":
+    m = math_model = MathModel()
+
+    m.variables["x1"] = FloatVar( 0.05, 2.0 )
+    m.variables["x2"] = FloatVar( 0.25, 1.3 )
+    m.variables["x3"] = FloatVar( 2.0, 15.0 )
+
+    # optimum objective in https://ceur-ws.org/Vol-2255/paper2.pdf is 0.0126652327883
+    m.objectives["f"] = Objective("minimize", lambda v, u: (v["x3"] + 2)*v["x2"]*(v["x1"]**2))
+
+    m.constraints["g1"] = Constraint( lambda v, u: 1 - v["x2"]**3*v["x3"]/(71785*v["x1"]**4), "<=")
+    # in the article about Aquila authors lost "-1" in the end of "g2" constraint...
+    m.constraints["g2"] = Constraint( lambda v, u: (4*v["x2"]**2 - v["x1"]*v["x2"]) / (12566*(v["x2"]*v["x1"]**3 - v["x1"]**4)) + (1/(5108*v["x1"]**2)) - 1, "<=")
+    m.constraints["g3"] = Constraint( lambda v, u: 1 - (140.45*v["x1"])/(v["x2"]**2*v["x3"]), "<=")
+    m.constraints["g4"] = Constraint( lambda v, u: (v["x1"] + v["x2"])/1.5 - 1, "<=")
+
+    termination_strategy = ScoreNoImprovement(time_seconds_limit=5)
+    agent = GeneticAlgorithm(population_size=256, crossover_probability=0.8, p_best_rate=0.5,
+                             tabu_entity_rate=0.2, mutation_rate_multiplier=1.0, move_probas=[0.5, 0.5, 0.0, 0.0, 0.0, 0.0],
+                             migration_rate=0.00001, migration_frequency=10, termination_strategy=termination_strategy)
+    
+    solver = SolverPureMath(math_model, agent,
+                            ParallelizationBackend.Multiprocessing, LoggingLevel.FreshOnly,
+                            n_jobs=10, score_precision=[4, 2])
+    solution = solver.solve()
+
+    m.variables["x1"] = solution.variable_values_dict["x1"]
+    m.variables["x2"] = solution.variable_values_dict["x2"]
+    m.variables["x3"] = solution.variable_values_dict["x3"]
+    print( "Optimal values from GreyJack" )
+    math_model.explain_solution( solution )
+    print()
+
+    # optimal values from Aquila article
+    solution.variable_values_dict["x1"] = 0.0502439
+    solution.variable_values_dict["x2"] = 0.35262
+    solution.variable_values_dict["x3"] = 10.5425
+    print( "Optimal values from article about Aquila" )
+    math_model.explain_solution( solution )
+    print()
+
+    # optimal values from https://ceur-ws.org/Vol-2255/paper2.pdf
+    solution.variable_values_dict["x1"] = 0.051689156131
+    solution.variable_values_dict["x2"] = 0.356720026419
+    solution.variable_values_dict["x3"] = 11.288831695483
+    print( "Optimal values from https://ceur-ws.org/Vol-2255/paper2.pdf" )
+    math_model.explain_solution( solution )
+
+    print( "done" )