Artificial neural networks Deep learning

artificial neural networks (anns) or connectionist systems computing systems inspired biological neural networks constitute animal brains. such systems learn (progressively improve ability) tasks considering examples, without task-specific programming. example, in image recognition, might learn identify images contain cats analyzing example images have been manually labeled cat or no cat , using analytic results identify cats in other images. have found use in applications difficult express traditional computer algorithm using rule-based programming.

an ann based on collection of connected units called artificial neurons, (analogous axons in biological brain). each connection (synapse) between neurons can transmit signal neuron. receiving (postsynaptic) neuron can process signal(s) , signal downstream neurons connected it. neurons may have state, represented real numbers, typically between 0 , 1. neurons , synapses may have weight varies learning proceeds, can increase or decrease strength of signal sends downstream.

typically, neurons organized in layers. different layers may perform different kinds of transformations on inputs. signals travel first (input), last (output) layer, possibly after traversing layers multiple times.

the original goal of neural network approach solve problems in same way human brain would. on time, attention focused on matching specific mental abilities, leading deviations biology such backpropagation, or passing information in reverse direction , adjusting network reflect information.

neural networks have been used on variety of tasks, including computer vision, speech recognition, machine translation, social network filtering, playing board , video games , medical diagnosis.

as of 2017, neural networks typically have few thousand few million units , millions of connections. despite number being several order of magnitude less number of neurons on human brain, these networks can perform many tasks @ level beyond of humans (e.g., recognizing faces, playing go ).