Complexity measures in Genetic Programming learning: A brief review

Le, Nam; Xuan, Hoai Nguyen; Brabazon, Anthony; Thi, Thuong Pham

doi:10.1109/cec.2016.7744087

Cited by 25 publications

(13 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The overall GA used in the experiments is given as follows: Since in this work we are searching for interpretable solutions, a measure of complexity has to be established. Measuring the complexity of an individual can generally be stated with respect to its genotype (structural) or with respect to its phenotype (functional) (Le, Xuan, Brabazon, and Thi, 2016). Here, we decided to use a simple nodecounting measuring strategy where different types of functions, variables and terminals are counted with different weightings.…”

Section: Genetic Programmingmentioning

confidence: 99%

Interpretable policies for reinforcement learning by genetic programming

Hein

Udluft

Runkler

2018

Engineering Applications of Artificial Intelligence

105

View full text Add to dashboard Cite

The search for interpretable reinforcement learning policies is of high academic and industrial interest. Especially for industrial systems, domain experts are more likely to deploy autonomously learned controllers if they are understandable and convenient to evaluate. Basic algebraic equations are supposed to meet these requirements, as long as they are restricted to an adequate complexity. Here we introduce the genetic programming for reinforcement learning (GPRL) approach based on model-based batch reinforcement learning and genetic programming, which autonomously learns policy equations from pre-existing default state-action trajectory samples. GPRL is compared to a straight-forward method which utilizes genetic programming for symbolic regression, yielding policies imitating an existing well-performing, but non-interpretable policy. Experiments on three reinforcement learning benchmarks, i.e., mountain car, cart-pole balancing, and industrial benchmark, demonstrate the superiority of our GPRL approach compared to the symbolic regression method. GPRL is capable of producing well-performing interpretable reinforcement learning policies from pre-existing default trajectory data.

show abstract

Section: Genetic Programmingmentioning

confidence: 99%

Interpretable policies for reinforcement learning by genetic programming

Hein

Udluft

Runkler

2018

Engineering Applications of Artificial Intelligence

105

View full text Add to dashboard Cite

show abstract

“…In the context of empirical modeling using GP, Le et al [30] have recently reviewed the use of complexity measures, and point out the critical importance of trading off goodness-of-fit to the training data against model complexity; see also [36,Chap 7]. To explicitly address this trade-off here, we have used a global multiobjective GP formulation in this work with conventional tree-based individuals where the single population was sorted according to Pareto dominance.…”

Section: Evolutionary Frameworkmentioning

confidence: 99%

“…The principal contribution of this paper is an investigation of the performance of GP approaches when supplemented by semantically-aware local search methods. In particular, this paper extends consideration of the effectiveness of local search to a multiobjective (MO) GP framework since this explicitly trades off goodness-of-fit against model complexity, a key requirement in the empirical modeling of data [10,30]. For the reasons stated in the preceding paragraph, we also make comparison with the RDO approach.…”

Section: Introductionmentioning

confidence: 99%

Comparison of semantic-based local search methods for multiobjective genetic programming

Dou

Rockett

2018

Genet Program Evolvable Mach

View full text Add to dashboard Cite

We report a series of experiments that use semantic-based local search within a multiobjective genetic programming (GP) framework. We compare various ways of selecting target subtrees for local search as well as different methods for performing that search; we have also made comparison with the random desired operator of Pawlak et al. using statistical hypothesis testing. We find that a standard steady state or generational GP followed by a carefully-designed single-objective GP implementing semantic-based local search produces models that are mode accurate and with statistically smaller (or equal) tree size than those generated by the corresponding baseline GP algorithms. The depth fair selection strategy of Ito et al. is found to perform best compared with other subtree selection methods in the model refinement.

show abstract

“…Genetic programming, which provides a novel way for efficient cache replacement policy, is an evolutionary methodology whereas an alternative to a solution; the population involves programs . The programs are typically a naive illustration of a tree, graph, or set of command (instructions).…”

Section: The Proposed Intelligent Cache Replacement Modelmentioning

confidence: 99%

“…Therefore, an approximate solution is acceptable, so GP is a good solution for evolving a good cache replacement strategy. Genetic programming is an evolutionary computation procedure that by design dissolves difficulties devoid of demanding the handler to recognize or identify the formula or construction of the solution beforehand . Genetic programming organizes this by arbitrarily producing a population of computer programs (demonstrated by tree context) and then mutating and crossing over the finest carrying out trees to generate a novel population.…”

Section: Introductionmentioning

confidence: 99%

An intelligent database proactive cache replacement policy for mobile communication system based on genetic programming

Darwish

El-Shnawy

2018

Int J Communication

View full text Add to dashboard Cite

In the mobile environment, the movement of the users, disconnected modes, many data updates, power battery consumption, limited cache size, and limited bandwidth impose significant challenges in information access. Caching is considered one of the most important concepts to deal with these challenges.There are 2 general topics related to the client cache policy: cache invalidation method keeps data in the cache up to date and cache replacement method chooses the cached element(s) that would be removed from the cache once the cache stays full. The aim of this work is to introduce a new technique for cache replacement in a mobile database that takes into consideration the impact of invalidation time for enhancing data availability in the mobile environment by using genetic programming. In this case, each client collects information for every cached item in the cache like access probability, cached document size, and validation time and uses these factors in a fitness function to determine cached items that will be removed from the cache. The experiments were carried by NS2 simulator to assess the efficiency of the proposed method, and the outcomes are judged against existing cache replacement algorithms. It is concluded that the proposed approach performs significantly better than other approaches. KEYWORDScache invalidation, cache replacement, genetic programming, mobile database | INTRODUCTIONWith the rapid changes in technology, mobile communications are the essential part of our daily life. People are using the wireless network for making phone calls, accessing the Internet to download news, the stock market, etc. Different techniques for enhancing the capability and quality of wireless environments have become a hot research area. The development of mobile information systems has met some disadvantages. Mobile machines regularly have slight storage dimension and inadequate energy. Furthermore, mobile client users in wireless environments suffer from frequent disconnections from the server, and also calling the server for information is costly. For that reason, scholars have discovered caching approaches for valuable data managing in the mobile podium. 1-3 Still, a key difficulty is how to preserve the collected items in the cache up to date with the source data at the server.

show abstract

Complexity measures in Genetic Programming learning: A brief review

Cited by 25 publications

References 42 publications

Interpretable policies for reinforcement learning by genetic programming

Interpretable policies for reinforcement learning by genetic programming

Comparison of semantic-based local search methods for multiobjective genetic programming

An intelligent database proactive cache replacement policy for mobile communication system based on genetic programming

Contact Info

Product

Resources

About