Deep Learning for Edge Computing: A Survey

A compendium of Deep Learning algorithms that are applied in Edge Computing through IoT devices that generate a large amount of data and representative image types was analyzed. The problem is the lack of proposals for the analysis and recommendation of Deep Learning algorithms that use data from devices in an Edge Computing. The objective is to carry out a survey as a research technique for the collection of data on Deep Learning algorithms for Edge Computing in different application areas, and determine which algorithm is most used and efficient. The applied methodology is exploratory research for the study of problems associated with the treatment and processing of data in different areas, gradual approach and deduction based on information from selected references on Deep Learning and Edge Computing. This research resulted in an Impact of Deep Learning on Edge Computing, a Deep Learning Algorithm Summary and an Edge Computing Summary resulted. It was concluded that according to the analysis carried out, the best algorithm is DNN due to its high degree of precision 98% in different areas; DNN stands out for the complexity of its structure that uses multiple layers (input, hidden and output) for the processing and training of the neurons that make it up.