SeaPearl.jl

Julia hybrid constraint programming solver enhanced by a reinforcement learning driven search.
Popularity
168 Stars
Updated Last
28 Days Ago
Started In
May 2020

codecov license Docs Build TagBot

drawing

SeaPearl is a Constraint Programming solver that can use Reinforcement Learning agents as value-selection heuristics, using graphs as inputs for the agent's approximator. It is to be seen as a tool for researchers that gives the possibility to go above and beyond what has already been done with it.

The paper accompanying this solver can be found on the arXiv. If you use SeaPearl in your research, please cite our work.

The RL agents are defined using ReinforcementLearning.jl, their inputs are dealt with using Flux.jl. The CP part, inspired from MiniCP, is focused on readability. The code is meant to be clear and modular so that researchers could easily get access to CP data and use it as input for their ML model.

Installation

]add SeaPearl

Use

Working examples can be found in SeaPearlZoo and documentation can be found here.

SeaPearl can be used either as a classic CP solver that uses predefined variable and value selection heuristics or as Reinforcement Learning driven CP solver that is capable of learning through solving automatically generated instances of a given problem (knapsack, tsptw, graphcoloring, EternityII ...).

SeaPearl as a classic CP solver :

To use SeaPearl as a classic CP solver, one needs to :

  1. declare a variable selection heuristic :
YourVariableSelectionHeuristic{TakeObjective} <: SeaPearl.AbstractVariableSelection{TakeObjective}
  1. declare a value selection heuristic :
BasicHeuristic <: ValueSelection
  1. create a Constraint Programming Model :
trailer = SeaPearl.Trailer()
model = SeaPearl.CPModel(trailer)

#create variable : 
SeaPearl.addVariable!(...)

#add constraints : 
SeaPearl.addConstraint!(model, SeaPearl.AbstractConstraint(...))

#add optionnal objective function : 
SeaPearl.addObjective!(model, ObjectiveVar)

SeaPearl as a RL-driven CP solver :

To use SeaPearl as a RL-driven CP solver, one needs to :

  1. declare a variable selection heuristic :
CustomVariableSelectionHeuristic{TakeObjective} <: SeaPearl.AbstractVariableSelection{TakeObjective}
  1. declare a value selection learnedheuristic :
LearnedHeuristic{SR<:AbstractStateRepresentation, R<:AbstractReward, A<:ActionOutput} <: ValueSelection
  1. define an agent :
agent = RL.Agent(
policy=(...),
trajectory=(...),
)
  1. optionally, declare a custom reward :
CustomReward <: SeaPearl.AbstractReward 
  1. optionally, declare a custom StateRepresentation ( instead of the Default tripartite-graph representation ) :
CustomStateRepresentation <: SeaPearl.AbstractStateRepresentation
  1. optionally, declare a custom featurization for the StateRepresentation :
CustomFeaturization <: SeaPearl.AbstractFeaturization
  1. create a generator for your given problem, that will create different instances of the specific problem used during the learning process.
CustomProblemGenerator <: AbstractModelGenerator
  1. set a number of training epochs, declare an evaluator, a Strategy, a metric for benchmarking
nb_epochs = 3000
CustomStrategy <: SearchStrategy #DFS, RBS, ILDS 

CustomEvaluator <: AbstractEvaluator #or use predefined one : SeaPearl.SameInstancesEvaluator(...)
function CustomMetricsFun
  1. launch the training :
metricsArray, eval_metricsArray = SeaPearl.train!(
valueSelectionArray=valueSelectionArray,
generator=tsptw_generator,
nbEpisodes=nbEpisodes,
strategy=strategy,
eval_strategy=eval_strategy,
variableHeuristic=variableSelection,
out_solver = true,
verbose = true,
evaluator=SeaPearl.SameInstancesEvaluator(valueSelectionArray,tsptw_generator; evalFreq = evalFreq, nbInstances = nbInstances, evalTimeOut = evalTimeOut),
restartPerInstances = restartPerInstances
)

Contributing to SeaPearl

All contributions are welcome! Have a look at our contributing guidelines.