| Автор | Maurizio Biasini |
| Дата выпуска | 1998-11-23 |
| dc.description | Preliminary results of an artificial neural network approach to 3D image reconstruction are reported. The set of (n - 1)-dimensional projections of the n-dimensional image and the pixels of the image itself constitute the input and output layers of the system, respectively. The layers are directly interconnected via a network of synapses, each associated with a specific weight. The network is trained using a supervised scheme: the projections of a set of point patterns are fed into the network and the reconstructed images are compared to the known patterns. The weights, organized as a matrix, are iteratively updated to minimize the mean square error between the ideal and reconstructed patterns, until convergence is reached. Once the training process is completed, the network can generalize and reconstruct any 3D density, one plane at a time, from the one-dimensional slices of the 2D experimental data. In this work, the network is trained to simultaneously deconvolute the ideal data from the experimental resolution. The symmetry of the momentum density is inserted in the learning procedure to reduce the memory requirements of the algorithm. The reconstructed data are compared with those obtained from the standard filtered-back-projection method. The numerical simulations considered show a reduction in mean square error of up to a factor of five compared to the back-projection results. |
| Формат | application.pdf |
| Издатель | Institute of Physics Publishing |
| Название | Three-dimensional reconstruction of the electron-positron momentum density via supervised artificial neural network |
| Тип | paper |
| DOI | 10.1088/0953-8984/10/46/018 |
| Electronic ISSN | 1361-648X |
| Print ISSN | 0953-8984 |
| Журнал | Journal of Physics: Condensed Matter |
| Том | 10 |
| Первая страница | 10517 |
| Последняя страница | 10528 |
| Аффилиация | Maurizio Biasini; ENEA (Istituto Nazionale di Fisica della Materia), via don Fiammelli 2, 40128 Bologna, Italy |
| Выпуск | 46 |
