All Classes and Interfaces
Class
Description
Represents a game position in an artificial game tree.
A class that stores the settings for a
ArtificialGamePosition tree.This class implements B* with a few additional modifications:
if the number of back-ups without leading back to the root reaches 1000, the algorithm is forced to back-up to the root
if a terminal node is selected, the algorithm is forced to back-up to the root
if the current node's best child is not its least pessimistic child, there is a 30% chance that the least pessimistic child is selected,
this will not occur if this would select a terminal node
Disprove-Best B*, as described in the report https://github.com/Pascalos99/ReportNewAdvancesInBstar/blob/main/Anema2025.pdf
A B*² implementation which uses Disprove-Best B* at the first-level search only.
This implements B*²-logistic with the `dispose all' bounds preservation technique and no shallow or deep irrelevance.
This implements B*²-max with the `dispose all' bounds preservation technique and no shallow or deep irrelevance.
Deprecated.
Large class containing many methods for running the experiments and creating new experiments.
Class for configuring the setup of algorithms to experiments for comparing them.
A simple debug class.
GameTreeNode implementation that does not store children nodes, but does store the depth of bounds and the bounds themselves.
An optional variant of the DepthFirstNode which stores the nodes at depth 1 in memory.
Deprecated.
This is the df-B* implementation applied in the thesis.
This implementation of df-B* with initial node selection inspired by B* Disprove-Best,
but not exactly following it.
This implementation of df-B* applies concepts from B* Disprove-Best into the
internal node selection procedure.
Deprecated.
The first core experiment, with the goal of performing a general comparison of algorithm variants.
The second core experiment, with the goal of determining the exact memory footprint of B*-squared compared to B*
This is part of the preliminary experiments conducted to determine the value for
P_{alt} as chosen in the thesis, which was found to at best be P_{alt}=0.3
This class represents an abstract GameTreeNode which stores game positions
of a game tree along with the structure of the tree itself through
GameTreeNode.children(MetricKeeper...) and GameTreeNode.parent() relations.A record to hold the two best nodes (and their indices in the original collection).
Generates the tree figures included in the thesis.
A generator for customising the tree structure of artificial game trees made with the
VariantAGP class.The generator type specifically used for generating the bounds of the children
generated as the children of each node of a
VariantAGP artificial game tree.The generator type specifically used for generating the variable branching factor
at each node of a
VariantAGP artificial game tree.This interface represents a game position in any game which can be modelled as an adversarial or non-adversarial game.
This class is intended for use with algorithm testing by keeping
track of metrics used to evaluate different algorithms on game tree
search problems.
Creates a tree of position type Q from a source tree with position type P with the help of a P-to-Q transformer.
General interface for all search algorithms.
The limits imposed on the search.
Encapsulates the results and metrics captured from an adversarial tree search.
This represents a search algorithm which primarily stores search results in a transposition table.
This represents a search algorithm which can search a given tree and modify it during search.
Implementation of
GameTreeNode that stores all nodes directly in memory.A class that stores the settings for a
SearchTreeNode tree.Implementation of
GameTreeNode that only acts as a facade for the underlying
tree expressed by the P implementation of IGamePosition.A basic implementation of
IGamePosition which allows for adding of children to the tree and setting their values.A stop condition for B* search.
The heuristic function for determining the strategy at the root of any B* search.
A very minimal transposition table implementation for df-B*.
Deprecated.
This class is part of the preliminary experiments performed on df-B* to find the values for
sigma and epsilon chosen in the thesis for the core experiments.
This class allows the manual exploration of trees.
Represents a game position in a variant artificial game tree.
A record holding the information provided to
Generator functions for determining the bounds of a generated child node.A class that stores the settings for a
VariantAGP tree.