Evulotionary Optimization Flashcards
In evulotionary optimization the problem to be solved must be ___
well defined
The comparison between two or more candidate solutions are based on ___ of how well a proposed solution meets the ___ of the problem
quantitive measures
needs
Interactive evolutionary computation is when a human provides a judgement about ___
Here we usually find the use of ___
. the quality of proposed solutions
. qualitative or even fuzzy descriptors
In quantitative evolutionary optimization, is a ___ that operates on a potencial ___ and returns either a single real number or multiple numbers that describe the value of the solution
function solution
There are two forms of optimization problem:
1 - ___
2- ___
1- numeric
2- combinatoric
In a Global Numerical Optimization we: 1- \_\_\_ 2- \_\_\_ 3- \_\_\_ 4- \_\_\_ 5- \_\_\_
1- Initialize Population 2- Create Offspring 3- Score Everyone 4- Select new parents 5- repeat to 2 until done
In a Global Numerical Optimization we should stop if:
1- The best solution is below a ___
2- The best solution is not ___
1- pre-defined threshold
2- evolving for a pre-defined time (number of generations)
When we generate offsprings from mutation we apply a variation on each dimension and is based on only ___
one parent
When we generate offsprings from recombination, it is based on more ___ by using ___ and ___
. than one parent
. crossover and belnding
In one point crossover we basicly make a ___ on the parents ___
cut
vectors
In a multipoint crossover we basicly make several ___ on the parents ___
cuts
vectors
In a uniform crossover we select one ___ from either parent at ___ without regard to maintaining ___.
. component
. random
. continuous segment
In a Blending we can use for example a ___ to blend two parents
mean
If the problem at hand presents a smooth, convex, continuous landscape thengradient or related methods of optimization will be ___ in locatting the ___
. faster
. single optimum point
if the problem presents a landscape with
multiple local optima, then the gradient methods will likely fail to find the ___
global optimum
If the landscape is discontinuous and/or not
smooth, then gradient-based approaches
may be ___
inapplicable
STUDY SLIDE 19 to SLIDE 20 (inclusive)
…
Some mutation alternatives to create offsprings are: 1- \_\_\_ 2- \_\_\_ 3- \_\_\_ 4- \_\_\_
1- Select and replace
2- Invert
3- Protect and randomize
4- Partially mapped crossover
Convergence with probability 1 happens with an Elistic selection and a mutation operator that can ___
reach all possible states
Premature convergence occurs when a population becomes ___ at a solution that is not the ___
. homogeneous
. the global optimum
When selecting a representation it should:
1- Optimally provide ___ about the solution
2- Be ___ to ___ operators
1- immediate information
2- amenable / variation
Selection describes either:
1- The process of ___ solutions from an existing ___
2- Or making proportionally more ___ from certain ___
1- selecting/ population
2- offspring / parents
Some forms of selection are: 1- \_\_\_ 2- \_\_\_ 3- \_\_\_ 4- \_\_\_
1- plus/comma
2- proportional
3- tournament
4- linear ranking
Plus/Comma selection includes allowing ___ to ___ to a maximum of n generations
. parents
. survive
Proportional Selection or Roulette Wheel Selection picks parents for ___ in proportion to their relative ___
reproduction
fitness
In Plus selection both ___ and ___ are ___ of the next generation
parents and offsprings
parents
In Comma selection only ___ are considerer for parents in the next generation
offsprings
In Tournament Selection we selct a number of individuals to be ___
a member of the next generation
In Linear Ranking Selection we map individuals to selection ___ according to a ___
probabilities
prescribed formula
Elistic selection can be applied to ensure that the ___
best solution in a
population is retained in the next generation
The harder the selection, the faster the ___ can ___
better solutions
overtake the population
If we are using evolutionary algorithms that relie heavily on recombination, then ___ is required for the population to ___
diversification
search new solutions
Some types of variations are:
1- ___ Variation
2- ___ Operators
3- Variations on ___
1- Real-Valued
2- Multiparent Recombination
3- Variable-Length Structures
Real-Valued Variation uses ___ mutation operator
Gaussian
Some of the constraints of the problem are part of the ___, whereas some are part of the ___
objective
parameters of a solution
A Hard constraint is one that, if violated, makes the entire proposed solution ___
worthless
A Soft constraint is one that can be violated, but there is some imposed ___. This can increase with the degree of ___
penalty
violation
When thinking about constraint handling, it is helpfull to craft an objective function that involves the primary criteria of interest and also involves a ___ that treats constraint violations
penaly function
The 1/5 rule ilustrates that static parameters for variation operators are very unlikely to lead to the ___
best rates of progress
Meta-Evolution is a commom method that views the parameters that control the ___ as part of the ___
. evolutionary search
. evolutionary process
Meta-evolution can be extended to ___ that are not in the continuous domain.
support variation operators
Multiparent Recombination Operators (type of variation) recombines ___ or ___ of multiple solutions
elements or blend parameters
Variations on Variable-Length Structures (type of variation) represents the variation on ___
data structures of variable length
