✔
Evolutionary computing
Created @September 7, 2021 10:01 AM
Class
Type
Materials
Lecture 1
introduction
triangle of live:
Lecture 2
evolutionary problem solving
problem and problem instance are not the same
constructive method: starting with empty solution and extending it one by one
heuristic method: trying to minimize need for backtracks, educated than greedy (taking the first)
iterative improvement method: starting random and trying to make it better
Evolutionary computing; only keeping best ones (quality-based selection) after iterative and heuristic
approach → link between problem solving as a context and biological evolution
EC metaphor:
environment: problem
individuals: candidate solutions
natural selection based on fitness: quality
Evolutionary computing 1
, reproduction
EC toolkit:
evolvable objects-phenotypes, what do you want? Representing the problem in digital code →
genetic code-genotypes, can be mutated and crossover to generate new individuals
reproduction
fitness
selection
Lecture 3
chapter 1: problems to be solved
black box model
3 components: input, model, and output
when one component is unknown: new problem type
optimization: input unknown, for example a uni timetable
modelling: model is unknown, model has to give the correct output with the given input.
modelling problems can be transformed into optimization problems
Evolutionary computing 2
, simulation: output unknown, often used to answer "what if" questions in evolving dynamic
environments, for example impact analysis new tax systems or weather forecast systems
search problems
difference between problems and problem-solvers
search problems, which define search spaces, and
problems-solvers, which move through search spaces to find a solution
optimization vs constraint satisfaction
objective function: a way of assigning a value to a possible solution that reflects its quality on scale
constraints: binary evaluation telling whether a given requirement holds
constraint problems can be transformed into optimization problems
constrained is a noun optimization problem that is subject to constraint, constraint is that what needs
to be satisfied
NP problems
so far problem type was only depending on the problem only, now classification scheme by
looking at properties of the problem solver. Looking at the difficulty/hardness of the solving
problem
problem size: number of problem variables (dimensionality) and number of different values for the
problem variables
running time: number of operations the algo takes to terminate. worst case as a function of
problem-size
problem reduction: current problem → map → another problem (transformation), solution for the
other problem is also a solution for the current problem
hardness scheme
Evolutionary computing 3
, lecture 4
chapter 2: the origins
background
fathers of evolutionary computing: darwin, founders of genetics
motivation for evolutionary computing
if evolution can develop intelligence, than artificial evolution can develop artificial intelligence ⇒
high level
Evolutionary computing 4
Evolutionary computing
Created @September 7, 2021 10:01 AM
Class
Type
Materials
Lecture 1
introduction
triangle of live:
Lecture 2
evolutionary problem solving
problem and problem instance are not the same
constructive method: starting with empty solution and extending it one by one
heuristic method: trying to minimize need for backtracks, educated than greedy (taking the first)
iterative improvement method: starting random and trying to make it better
Evolutionary computing; only keeping best ones (quality-based selection) after iterative and heuristic
approach → link between problem solving as a context and biological evolution
EC metaphor:
environment: problem
individuals: candidate solutions
natural selection based on fitness: quality
Evolutionary computing 1
, reproduction
EC toolkit:
evolvable objects-phenotypes, what do you want? Representing the problem in digital code →
genetic code-genotypes, can be mutated and crossover to generate new individuals
reproduction
fitness
selection
Lecture 3
chapter 1: problems to be solved
black box model
3 components: input, model, and output
when one component is unknown: new problem type
optimization: input unknown, for example a uni timetable
modelling: model is unknown, model has to give the correct output with the given input.
modelling problems can be transformed into optimization problems
Evolutionary computing 2
, simulation: output unknown, often used to answer "what if" questions in evolving dynamic
environments, for example impact analysis new tax systems or weather forecast systems
search problems
difference between problems and problem-solvers
search problems, which define search spaces, and
problems-solvers, which move through search spaces to find a solution
optimization vs constraint satisfaction
objective function: a way of assigning a value to a possible solution that reflects its quality on scale
constraints: binary evaluation telling whether a given requirement holds
constraint problems can be transformed into optimization problems
constrained is a noun optimization problem that is subject to constraint, constraint is that what needs
to be satisfied
NP problems
so far problem type was only depending on the problem only, now classification scheme by
looking at properties of the problem solver. Looking at the difficulty/hardness of the solving
problem
problem size: number of problem variables (dimensionality) and number of different values for the
problem variables
running time: number of operations the algo takes to terminate. worst case as a function of
problem-size
problem reduction: current problem → map → another problem (transformation), solution for the
other problem is also a solution for the current problem
hardness scheme
Evolutionary computing 3
, lecture 4
chapter 2: the origins
background
fathers of evolutionary computing: darwin, founders of genetics
motivation for evolutionary computing
if evolution can develop intelligence, than artificial evolution can develop artificial intelligence ⇒
high level
Evolutionary computing 4
