Convolutional Neural Networks: The Possibilities are Almost Endless

Interest in modeling the workings of the brain has plagued our species for centuries. With the advent of neuroscience, this became possible as neuron structure was discovered and further studied. Researchers in the mid-20th century began formulating “artificial neural networks,” architectures that mimicked the human brain through the use of hidden layers and mathematical activation functions likened to neuronal activity. Results were academically interesting, but network scale was limited by computational processing power and the availabili ty of usable, annotated datasets for refining tasks like visual recognition.

Source: Current Problems in Diagnostic Radiology - February 21, 2018 Category: Radiology Authors: Rebecca J. Mieloszyk, Puneet Bhargava Source Type: research