Mkrtchyan F.A. The Problem of Learning to Make Statistical Decisions for Small Samples for Remote Monitoring Marine Ecosystems. In: PICES-2022 Annual Meeting, 23-30 September 2022, Busan, Korea , PICES , р. 18.
Полный текст не доступен из этого репозитория.Аннотация
Application of means of remote monitoring in many cases is connected with acceptance of the statistical decision on presence on a surveyed part of studied space of this or that phenomenon. One of features of conditions of gathering of the information for such decision is the impossibility of reception statistical samples great volumes. Therefore working out and research of optimum algorithms of distinction of the casual signals characterized by samples of the limited volume, in the conditions of parametrical aprioristic indefinite are necessary. The article considers the "Spotting" model as an informative parameter of the background characteristics of the studied space according to remote sensing data. The most obvious way to identify spots is the threshold setting method. In this case, the area of the spot includes that part of the space in which the indicator of the medium for this channel exceeds (l+- characteristic) or does not exceed (l-- characteristic) the threshold value. The work is carried out modular structure of the statistical modeling system of spotting The structure of the software is proposed. The analysis of empirical histograms for ''spottiness'' shows, that in most cases (l+, l-) - characteristics will be coordinated with exponential distribution, and amplitude characteristics will be coordinated with normal distribution. Therefore for detection and classification of the phenomena on a surface of ocean it is necessary to apply optimal algorithms for the Computer training to taking statistical decisions for the aforesaid distributions In the present work the generalized adaptive algorithm of training to acceptance of statistical decisions for exponential classes of distributions is developed at aprioristic parametrical uncertainty of conditions small samples. Numerical examples are shown. Efficiency of the developed optimum procedure for small samples is shown.
| Тип объекта: | Доклад на конференции или семинаре (Доклад) |
|---|---|
| Авторы на русском. ОБЯЗАТЕЛЬНО ДЛЯ АНГЛОЯЗЫЧНЫХ ПУБЛИКАЦИЙ!: | Мкртчян Ф.А. |
| Подразделения (можно выбрать несколько, удерживая Ctrl): | 209 лаб. вычислительная |
| URI: | http://cplire.ru:8080/id/eprint/8229 |
 |
Изменить объект |
