Abstract
Due to the lack of high-end graphics or tensor processing units, previously, deep neural networks could not be implemented as state-of-the-art Artificial Intelligence (AI) algorithms. Rather, linear models were preferred, and they were easy to understand and interpret. Things started changing with the advent of more advanced processing units, in the last decade, when the algorithms took on real-world problems. The models began getting bigger and better.
Involved units
Leibniz-HKI-Authors
Identifier
doi: 10.1007/978-3-031-12807-3_6