v0.4.1

SeaPearl v0.4.1

Diff since v0.4.0

Closed issues:

• Tackle the "stochastic" issue while training: taking the difficulty of an episode into account. (#23)
• Explore the idea of having two (or more) agents to solve a problem (#30)
• Make it possible to give rewards later (not directly after an action is taken) (#32)
• Try to give some metrics as input of the RL agent (#34)
• Make the variable heuristic an RL agent learned heuristic as well (#35)
• Add other type of search strategy (#38)
• Support Max sense for @objective (#42)
• Make JuMP GraphColoring tests deterministic (#43)
• Representable variables (#72)
• TstptwReward weird behaviour with numberOfDecisions (#96)
• Set up documentation (#102)
• Testset for CPreward2 (#190)
• [CP] No easy way to generate a multiple of a variable, y = ax (#196)
• [CP] No easy way to generate an offsetted variable, y = x + c (#197)
• [CP] No easy way to generate the opposite of a variable, y = -x (#198)
• [CP][Boolean Operator] Implement truth-functional operator of logical conjunction : ∧ (#200)
• Remove dead code, useless/test comments and empty files (#201)
• TSPTW problem refactoring (#211)
• DefaultStateRepresentation improvement (#212)
• Add a random value selection heuristic. (#213)
• Generic reward encouraging smart exploration of the tree. (#216)
• Refacto of the Graph Convolution (#225)
• Implementation of the heterogeneous graph (#227)
• Adapt New heterogeneous Graph convolution to FullFeaturedCPNN and VariableOutputCPNN (#230)

Merged pull requests:

• Feature/disjonctive detectable precedence (#156) (@kimriouxparadis)
• Variant of CPReward (#189) (@tomsander1998)
• fix CPReward2 constructor and dependencies (#191) (@3rdCore)
• fix type in feature_length (#192) (@marco-novaes98)
• minor modification in sumlessthan.jl (#193) (@marco-novaes98)
• Add variablesArray for NotEqualConstant (#194) (@marco-novaes98)
• Added features for experiment reproductibility and many other improvement : (#203) (@3rdCore)
• Add new search strategy: Large Neighborhood Search (LNS) (#204) (@marco-novaes98)
• Make type of earliestStartingTime more abstract (#206) (@marco-novaes98)
• Add summary of the performance test (#207) (@marco-novaes98)
• Override the - operator to easily define constraints of the form y= -x. (#208) (@gostreap)
• Tackling issues #196 and #200 (#209) (@louis-gautier)
• Add BinaryXor and isBinaryXor constraint (#210) (@gostreap)
• Optimization of the CP solver (#214) (@marco-novaes98)
• Adds chosenFeatures for a more expressive DefaultStateRepresentation. (#215) (@gostreap)
• RB generator (#217) (@gostreap)
• Add RandomHeuristic for value (#218) (@gostreap)
• Add RandomHeuristic for value (#219) (@gostreap)
• RB generator (#220) (@gostreap)
• Adds chosenFeatures for a more expressive DefaultStateRepresentation. (#221) (@gostreap)
• Add SolutionFound to BasicMetrics (#222) (@gostreap)
• Enable imitation learning (#223) (@ziadelassal)
• Added instance generator for KEP (#224) (@marco-novaes98)
• Creating a general reward encouraging a smart exploration of the tree (#226) (@louis-gautier)
• Small improvement in the way the helpSolution is retrieved in SupervisedLearnedHeuristic (#228) (@gostreap)
• Heterogeneous graph and new associated RL pipeline. (#229) (@gostreap)
• Maximum Independent Set Generator (#234) (@gostreap)
• Heterogeneous Graph Transformer (#235) (@louis-gautier)
• Correction of inconsistencies on GeneralReward (#237) (@louis-gautier)
• Max pooling for GraphConv (#238) (@3rdCore)
• New explorers and tripartite graph plot utils (#239) (@louis-gautier)
• Add erdos-renyi and barabasi-albert graph generator (#240) (@gostreap)
• Add a Max-Cut problem generator (#241) (@gostreap)
• Specific state representations for graph coloring and MIS (#242) (@louis-gautier)
• Lots of changed (GPU support, max constraint, dfswb ...) (#244) (@3rdCore)
• Added documentation automated deployment (#246) (@jardinetsouffleton)
• Added codecov to CI (#247) (@jardinetsouffleton)
• Draft: slight code cleanup: harmonized variable names and formatting (#252) (@jardinetsouffleton)
• Changed variable name to fix bug (#254) (@jardinetsouffleton)
• Revert "Draft: slight code cleanup: harmonized variable names and formatting" (#257) (@jardinetsouffleton)
• @JuliaRegistrator register (#266) (@jardinetsouffleton)
• Fixes for automated merge of package (#268) (@jardinetsouffleton)
• Fix/resolving compat issues (#270) (@jardinetsouffleton)
• Resolving compat format (#271) (@jardinetsouffleton)
• Resolving compat format (#272) (@jardinetsouffleton)
• setting julia version to 1.6-1.7 (#273) (@jardinetsouffleton)
• Removing dependencies: Revise, Statistics (#274) (@jardinetsouffleton)