J. E. Bowcock

Learn More
We propose here a nev method for retrieving an image from noisy, incomplete data. It is based on an idea incorporating the principle of maximum entropy, a veil known powerful method for image processing. An image is built on the computer by allocating grains to pixels one at a time on the basis of certain deterministic rules as to where each successive(More)
-We investigate a way in which deterministic networks can be trained to encode information using weights and biases by the method of simulated annealing. Results are presented for the simple XOR (exclusive-or) network which show that the current method of back-propagation can lead to false solutions, whereas simulated annealing leads reliably to true(More)
  • 1