DEAP/deap
Deap
Features
- Benchmarking Platforms - Provides a platform for prototyping, logging, and analyzing evolutionary algorithm performance with statistics tracking.
- Custom Evolutionary Components - Lets users create custom fitness functions, individual representations, and genetic operators from modular building blocks.
- Custom Genomes - Creates a user-defined data structure to represent an individual solution in an evolutionary algorithm.
- Evolutionary Algorithms - Provides a framework for defining custom genetic representations and operators to build tailored evolutionary algorithms.
- Genetic Algorithms - Provides a full-featured genetic algorithm library with selection, crossover, and mutation operators.
- Genetic Operator Implementations - Provides a library of selection, crossover, and mutation operators for evolving populations.
- Genetic Programming Toolkits - Ships a dedicated toolkit for evolving tree-based programs and symbolic expressions via genetic programming.
- Tree-Based Genetic Programming - Supports tree-based representations and operators for evolving programs and symbolic expressions.
- Operator Pipeline Frameworks - Defines evolution as a pipeline of composable operator objects chained together for the generation loop.
- Rapid Prototyping Frameworks - Provides a modular framework for rapidly prototyping evolutionary algorithms with custom components.
- Genetic Algorithms - Ships predefined algorithms for evolution strategies, genetic algorithms, and genetic programming.
- Evolutionary Statistics Loggers - Records and reports population metrics like fitness mean and standard deviation during an evolution run.
- Genetic Chromosome Types - Provides a flexible type system for defining custom chromosomes and individuals in evolutionary algorithms.
- Custom Constraint Functions - Applies custom constraint functions during evolution to ensure generated individuals satisfy domain-specific rules or bounds.
- Adaptive Mutation Strategies - Adapts the mutation step size or covariance matrix during evolution to improve convergence on real-valued problems.
- Genetic Program Evolution - Represents solutions as tree structures and evolves them using crossover, mutation, and selection.
- Genetic Program Trees - Represents programs as nested tree structures with typed nodes for genetic programming operations.
- Swarm Optimization Algorithms - Moves a population of candidate solutions through the search space using velocity updates informed by personal and global best positions.
- Artificial Ant Navigators - Evolves a control program that guides an agent to collect food tokens on a toroidal grid.
- Binary String Optimizers - Evolves a population of binary strings toward a target using selection, crossover, and mutation operators.
- Boolean Circuit Solvers - Evolves a program that correctly evaluates a truth table using genetic programming with boolean primitives.
- Parallel Evaluators - Distributes fitness evaluations across multiple processors or nodes using multiprocessing or SCOOP.
- Transparent Evolutionary Parallelizers - Distributes fitness evaluations across multiple cores or nodes transparently without changing algorithm code.
- Evolutionary Penalty Functions - Applies penalty functions to fitness evaluations to steer populations away from invalid solutions.
- Estimation of Distribution Algorithms - Supports estimation of distribution algorithms that learn and sample from probabilistic models of promising solutions.
- Evolutionary Parallelism Frameworks - Implements transparent parallel evaluation of fitness functions across multiple processors or nodes.
- State Checkpointing - Saves the entire evolutionary state to disk for exact resumption of runs.
- Transparent Parallel Evaluation - Wraps fitness evaluation in a parallel map for transparent multiprocessing without changing algorithm logic.
- Elite Solution Archives - Maintains a separate archive of elite individuals to preserve best solutions across generations.
- Generation Metric Logbooks - Collects per-generation metrics into a structured log for real-time and post-hoc analysis.
- Constraint Enforcement Decorators - Ships decorators that wrap fitness evaluation to enforce constraints and apply penalty functions.
- Resumable State Restoration - Serializes the full evolutionary state to disk for exact resumption of runs.
- Evolutionary Algorithms - Distributed evolutionary computation framework.
- General Machine Learning - Framework for evolutionary algorithms.
- Frameworks de Machine Learning - Framework for evolutionary algorithms in Python.
- Optimization - Evolutionary computation framework.
- Perception and Machine Learning - Framework for distributed evolutionary algorithms.
- Computation and Optimization - Framework for evolutionary computation and rapid prototyping.
