Representation and Hidden Bias: Gray vs. Binary Coding for Genetic Algorithms

Caruana, Richard A.; Schaffer, J. David

doi:10.1016/b978-0-934613-64-4.50021-9

Cited by 138 publications

(70 citation statements)

References 1 publication

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…In a large empirical study Schaffer et al have identified values for population size, crossover and mutation rate that produce good online performance on this test suite (Schaffer, et al, 1989); we are using their results (which are almost identical to earlier recommendations by John J. Grefenstette) here, and also follow their practice of Gray coding all genes (Caruana & Schaffer, 1988). To keep the experiment as straightforward as possible we did not follow the common practice of using an elitist strategy, 6 and kept the following parameters invariant unless noted otherwise: The search resolution was held constant at l = 3 for all experiments using DPE to facilitate comparison of results across the test suite; the control experiments used the values given in Table 1.…”

Section: Resultsmentioning

confidence: 68%

“…This disruptive effect of crossover complicates the above analysis considerably and would necessitate the introduction of a separate DPE trigger threshold (dependent on both crossover and mutation rate) for the center target interval. To avoid such undesirable complications the use of Gray code for DPE-mapped genes is highly recommended; it is also generally preferable as it reduces Hamming cliffs that could mislead the GA search (Caruana & Schaffer, 1988).…”

Section: Expected Convergence Levelmentioning

confidence: 99%

See 1 more Smart Citation

Dynamic Parameter Encoding for genetic algorithms

1992

View full text Add to dashboard Cite

Abstract. The common use of static binary place-value codes for real-valued parameters of the phenotype in Holland's genetic algorithm (GA) forces either the sacrifice of representational precision for efficiency of search or vice versa. Dynamic Parameter Encoding (DPE) is a mechanism that avoids this dilemma by using convergence statistics derived from the GA population to adaptively control the mapping from fixed-length binary genes to real values. DPE is shown to be empirically effective and amenable to analysis; we explore the problem of premature convergence in GAs through two convergence models.Keywords. Adaptive encoding, real-valued parameters, ARGOT, premature convergence, genetic hitchhiking MotivationGenetic Algorithms (GAs) are capable of efficiently searching complex fitness landscapes utilizing methods inspired by biological evolution (Holland, 1975). The work reported here used the GAuscd ~ software which implements a popular form of the GA, applying the operators of selective reproduction, mutation and crossover to a fixed-length binary genome. While the common restriction to fixed-length genetic representations in GA experiments is rooted in convention and convenience theoretical arguments have been made in support of binary alphabets (Goldberg, 1989, pp. 80-82).If the function to be optimized by the GA is defined over a continuous domain, its realvalued parameters can only have approximate genotypic representations as binary codes of a given length. The choice of code length determines not only the resolution of the GA search for this gene and the precision of its result, 2 but also the size of the space of representations the GA has to search.Unfortunately the GA searches long place-value codes rather inefficiently: in the early phases of the search a lot of computational effort is invested in evaluating the least significant digits of the gene. The optimal value for these digits, however, will typically depend on the more significant ones, and little useful information can be obtained from sampling the former until the latter have converged to their optimal value. Likewise, once this has happened there is no need to spend any more search effort on the most significant digits--the plain GA though will pay equal attention to all digits of the representation throughout its search, thus wasting its implicit parallelism 3 on an inherently sequential search problem.Moreover, in the absence of meaningful fitness information early in the search, the least significant bit positions will converge to an arbitrary value due to genetic drift, a process

show abstract

Section: Resultsmentioning

confidence: 68%

Section: Expected Convergence Levelmentioning

confidence: 99%

Dynamic Parameter Encoding for genetic algorithms

1992

View full text Add to dashboard Cite

show abstract

“…For example, neither "binary" nor gray coding (Caruana & Schaffer, 1988) can be said to be superior, either to each other or to any other representation, without reference to a particular problem domain. One immediate practical consequence of this should be a change of methodology regarding test suites of problems and comparative studies.…”

Section: Discussionmentioning

confidence: 99%

Fundamental limitations on search algorithms: Evolutionary computing in perspective

Radcliffe

Surry

1995

Computer Science Today

View full text Add to dashboard Cite

Abstract. The past twenty years has seen a rapid growth of interest in stochastic search algorithms, particularly those inspired by natural processes in physics and biology. Impressive results have been demonstrated on complex practical optimisation problems and related search applications taken from a variety of fields, but the theoretical understanding of these algorithms remains weak. This results partly from the insufficient attention that has been paid to results showing certain fundamental limitations on universal search algorithms, including the so-called "No Free Lunch" Theorem. This paper extends these results and draws out some of their implications for the design of search algorithms, and for the construction of useful representations. The resulting insights focus attention on tailoring algorithms and representations to particular problem classes by exploiting domain knowledge. This highlights the fundamental importance of gaining a better theoretical grasp of the ways in which such knowledge may be systematically exploited as a major research agenda for the future.

show abstract

“…As mentioned in Section 4.1, we use the binary genotype. For this encode, from the phenotype to genotype, we do not use binary-coding but Gray-coding, since Gray-coding is generally more superior (33) . We choose uniform crossover, because of its many advantages (34) .…”

Section: Ga Settingsmentioning

confidence: 99%

High Speed Genetic Lips Detection by Dynamic Search Domain Control

et al. 2007

View full text Add to dashboard Cite

In this paper, high-speed size and orientation invariant lips detection of a talking person in an active scene using template matching and genetic algorithms is proposed (refer to Fig. 1). As part of the objectives, we also try to acquire numerical parameters to represent the lips. The information is very important for many applications, where high performance is required, such as audio-visual speech recognition, and personal mobile devices interfaces. The difficulty in lips detection is mainly due to deformations and geometric changes of the lips during speech and the active scene by free camera motion. In order to enhance the performance in speed and accuracy, initially, the performance is improved on a single still image, that is, the base of video processing.Our proposed system is based on template matching using genetic algorithms (GA). Only one template is prepared per experiment. The template is the closed mouth of a subject, because the application is for personal devices. In our previous study, the main problem was trade-off between search accuracy and search speed. To overcome this problem, we use two methods: scaling window and dynamic search domain control (SD-Control).The transition of the objective value on the proposed system and the no SD-Control are plotted in Fig. 2. At points from (a) to (d) in Figs. 2, the transition of GA exploration is exactly the same in both cases. The reason is that the same random numbers are used in both cases. Between points (d) and (e) in Fig. 2, the differences in the methods appear. This is because, the search domain is controlled from point (d) in Fig. 2. The elite individual of the no SD-Control do not change from point (c) to (g), hence it is concluded that the GA exploration gets trapped into a local optimum. The reason why the exploration was trapped into a local optimum is that the small population loses the population diversity and causes premature convergence.On the other hand, in case of the proposed system, the evolution make good progress from generation = 117 (point from (e) to (i)). This indicates that the GA of the proposed system can escape from the local maximum. The reason for this is that a search domain is reduced by the dynamic SD-Control and the GA can explore the reduced search domain with meticulous detail. This means that the GA in the proposed system performs not only global optimization but also local optimization. The result of our demonstration clearly shows that a trade-off between exploration accuracy and speed is overcome by the dynamic SD-Control.We achieved a lips detection accuracy of 91.33% at an average processing time of 33.70 milliseconds per frame. Stephen Karungaru * * Non-member Minoru Fukumi * * MemberIn this paper, high-speed size and orientation invariant lips detection of a talking person in an active scene using template matching and genetic algorithms is proposed. As part of the objectives, we also try to acquire numerical parameters to represent the lips. The information is very important for many applications, where ...

show abstract

Representation and Hidden Bias: Gray vs. Binary Coding for Genetic Algorithms

Cited by 138 publications

References 1 publication

Dynamic Parameter Encoding for genetic algorithms

Dynamic Parameter Encoding for genetic algorithms

Fundamental limitations on search algorithms: Evolutionary computing in perspective

High Speed Genetic Lips Detection by Dynamic Search Domain Control

Contact Info

Product

Resources

About