Abstract
Using hierarchical neural network architectures (HNN), the authors study the implementations of several vision tasks, including contour finding, motion detection, and line detection (Hough transform). Several models are used. They are homogeneous and heterogeneous Boltzmann machines, homogeneous and heterogeneous Hopfield models, homogeneous winner-take-all model, and the models with and without limited displacement. Simulations have shown that the homogeneous system is capable of handling the desired vision tasks and thus simplifies the implementation.
Original language | English |
---|---|
Pages (from-to) | 511 |
Number of pages | 1 |
Journal | Neural Networks |
Volume | 1 |
Issue number | 1 SUPPL |
DOIs | |
Publication status | Published - 1988 |
Event | International Neural Network Society 1988 First Annual Meeting - Boston, MA, USA Duration: 1988 Sept 6 → 1988 Sept 10 |
All Science Journal Classification (ASJC) codes
- Cognitive Neuroscience
- Artificial Intelligence