A Modified Perception Algorithm for Computer-Assisted Diagnosis

摘要: We present a new stochastic learning algorithm and first results of computational experiments on fragments liver CT images. The is basically an extension the Perceptron by special type simulated annealing. images are size 119 x with 8 bit grey levels. From 220 positive (focal tumours) negative examples number hypotheses w 1 · + ··· n ≥ ϑ were calculated for = 14161 then tested voting function various sets 50 additional examples, respectively. input to derived from DICOM standard representation annealing procedure employs logarithmic cooling schedule c(k) Γ / ln (k 2), where parameter that depends underlying configuration space. In our experiments, chosen according estimations maximum escape depth local minima associated energy landscape.