Deep Learning -- A first Meta-Survey of selected Reviews across\n Scientific Disciplines, their Commonalities, Challenges and Research Impact

Abstract

Deep learning belongs to the field of artificial intelligence, where machines\nperform tasks that typically require some kind of human intelligence. Similar\nto the basic structure of a brain, a deep learning algorithm consists of an\nartificial neural network, which resembles the biological brain structure.\nMimicking the learning process of humans with their senses, deep learning\nnetworks are fed with (sensory) data, like texts, images, videos or sounds.\nThese networks outperform the state-of-the-art methods in different tasks and,\nbecause of this, the whole field saw an exponential growth during the last\nyears. This growth resulted in way over 10,000 publications per year in the\nlast years. For example, the search engine PubMed alone, which covers only a\nsub-set of all publications in the medical field, provides already over 11,000\nresults in Q3 2020 for the search term 'deep learning', and around 90% of these\nresults are from the last three years. Consequently, a complete overview over\nthe field of deep learning is already impossible to obtain and, in the near\nfuture, it will potentially become difficult to obtain an overview over a\nsubfield. However, there are several review articles about deep learning, which\nare focused on specific scientific fields or applications, for example deep\nlearning advances in computer vision or in specific tasks like object\ndetection. With these surveys as a foundation, the aim of this contribution is\nto provide a first high-level, categorized meta-survey of selected reviews on\ndeep learning across different scientific disciplines. The categories (computer\nvision, language processing, medical informatics and additional works) have\nbeen chosen according to the underlying data sources (image, language, medical,\nmixed). In addition, we review the common architectures, methods, pros, cons,\nevaluations, challenges and future directions for every sub-category.\n