As can be read in the webpage of the metaheuristics network (http://www.metaheuristics.org), a metaheuristic is a set of concepts that can be used to define heuristic methods that can be applied to a wide set of different problems. In other words, a metaheuristic can be seen as a general algorithmic framework which can be applied to different optimization problems with relatively few modifications to make them adapted to a specific problem. Examples of metaheuristics include evolutionary algorithms, simulated annealing, tabu search, iterated local search, and ant colony optimization. Metaheuristics have been widely used to solve different combinatorial (and numerical) optimization problems, with the goal of obtaining a very good solution (but perhaps not the optimum) to NP-complete problems in which exact search methods are intractable even for small problem sizes.
Dear Maysam,
A metaheuristic is an optimization algorithm aimed at solving difficult optimization problems (often coming from the fields of operations research, engineering or artificial intelligence) for which no more efficient classical method is known.
Metaheuristics are often inspired by natural systems, whether taken in physics (simulated annealing), in evolutionary biology (in the case of genetic algorithms) or in ethology (in the case of ant colony or particle swarm optimization).
Data Mining is in fact a generic term encompassing a whole family of tools facilitating the exploration and analysis of data contained within a Data Warehouse or DataMart decision base. The techniques put into action when using this analysis and prospecting instrument are particularly effective in extracting significant information from large amounts of data.
Metaheuristics and Data Minig, both techniques can use common problem solving algorithms such as Genetic Algorithm, Ant Colony Algorithms or Particle Swarm Optimization when solving each problem.
Best regards
Hi Maysam,
IMO, meta-heuristics and data mining are two groups of methods for supporting decisions in complicated problems. Nowadays, you can find heuristics in data mining approaches; and you can also find data mining method in modern heuristics.
1. Heuristics for data mining
The training phase of many -- if not all -- data mining models is actually an optimization problem:
argmaxx f(x).
where the objectives (f) usually consists of precision and recall ratio and the set of variables x may include a lot of parameters to tune. An optimization algorithm, such as a meta-heuristic, can help find the optimized model(s) quickly.
Examples are the stochastic gradient descent, Momentum, and Adam algorithms for training deep neural networks.
2. Data mining for heuristics
The no free lunch (NFL) theorems proved that no algorithm is statistically better than any other if solving ALL problems. Therefore, modern heuristics (hyper-heuristics and meta-heuristics) incorporate offline or online data mining to learn domain-specific, data-set-specific, or instance-specific features and adapt to the problem-solving. In other words, there can exist a higher level abstract problem
argmaxx g(x)
for an ordinary optimization problem
min f(x).
In this cases, data mining methods can discover the features and approximately discriminate scenarios for an advanced heuristic algorithm.
References
https://en.wikipedia.org/wiki/Stochastic_gradient_descent
https://www.youtube.com/watch?v=ILsA4nyG7I0
https://en.wikipedia.org/wiki/No_free_lunch_theorem
https://en.wikipedia.org/wiki/Hyper-heuristic
Hello,
The interaction between metaheuristics and data mining techniques seems to be a profitable marriage in order to enhance the results when we are trying to solve a problem related to one/both of the cited domains.
From a metaheuristic point of view, what we need is to be more intelligent by identifiying some promising parts of the search space. And this is can be provided by data mining techniques.
From a data mining point of view, what we need is to be more efficient/fast to find a solution or to converge. And this can be done by using metaheuristics.
best regards,
- Badges
- Science topic
- Similar topics
- Social Science
- Media
- Social Media