خط مشی دسترسیدرباره ما
ثبت نامثبت نام
راهنماراهنما
فارسی
ورودورود
صفحه اصلیصفحه اصلی
جستجوی مدارک
تمام متن
منابع دیجیتالی
رکورد قبلیرکورد بعدی
Record identifier : 569807
Personal Name - Primary Intelectual Responsibility : Zhang, Ying Yuan
Title and statement of responsibility : Point pattern reconstruction using significantly incomplete interpoint distance information: Multidimensional scaling and genetic algorithms approaches [Thesis]
Publication, Distribution,Etc. : Tufts University, 1998
Language of the Item : eng
Dissertation of thesis details and type of degree : Ph.D
Body granting the degree : , Tufts University
Summary or Abstract : This point pattern reconstruction problem is of NP hard. We deal with it as an optimal search over an unknown multimodal hypersurface. Two optimization routines, multidimensional scaling (MDS) and genetic algorithms (GA's) with different implementation techniques are used in this study for a 10-point two-dimensional pattern with numbers of interpoint distances gradually reduced from complete to theoretical GLB of minimum. Many interesting problems arose as the reconstruction information is reduced. This study shows that the response function corresponding to the incomplete reconstruction information becomes increasingly multimodal; the correlation between partial stress and full stress decreases; the difference between the satisfactory reconstruction in image recognition sense and the mathematical "near optimal solution" increase. Representation of point patterns by interpoint distances has solid practical background. It is preferable to coordinate representation under many circumstances. Point patterns are of interest in a number of contexts. Star constellations and molecular essentially consist of points. Complex shapes are often approximated by a relatively small number of "feature points". Many problems benefit from representing a pattern with as little information as is practical. These include issues of computer usage, reducing data storage and computational requirements, the time and cost of data acquisition and, in psychology experiments, the additional consideration of the resistance of human subjects to excessive testing. This study also finds that: GA's do explore greater region in search than MDS and efficiently converge to lots of near-optimal solutions. The important issue is to distinguish the global minimum from other local minima when low amounts of information are available. Multiple searches (MDS or GA's) can significantly improve the ability of search as well as recognize the global minimum solution. Hybrid techniques with GA's and MDS not only improving the efficiency of GA's but also greatly avoid GA's "premature convergence". There is no simple relationship between near optimal solution (in the mathematical sense) and "satisfactory solution" (in image recognition sense). What a subject judges to be a satisfactory is not only dependent on the value of partial stress but also related to the distribution of the partial stress. This study indicates that among different solutions with similar partial stress, the one with even stress distribution have the higher probability of being satisfactory solutions..
Topical Name Used as Subject : Operations research
: Systems design
: Computer science
Information of biblio record : TL
 
 
 
(در صورت عدم وضوح تصویر اینجا را کلیک نمایید)