Artificial Intelligence Based on Modular Reinforcement Learning and Unsupervised Learning
Shawan Taha Mohammed, Angelo Mihaltan, Kemal Acar, Hendrik Laux, Gerd Ascheid · 2021
The progress observed on simple multilayer perceptrons regarding convolutional neural networks and recurrent neural networks can be an accurate example in this manner. The deeper the networks become, the more difficult it is to optimize the millions of parameters by just sparse rewards. Reducing the parameters that are optimized via RL can make the problem much easier. Machine Learning is a sub-area of artificial intelligence. It is a data-driven method and aims to recognize patterns in data. Data is transferred to the model, then a prediction is made by it. The obtained output of the model is compared with the expected output, known as the ground truth. The difference between the prediction of the model and the ground truth is a cost with which the model is afterwards trained. In the course of the development in the last decade, it was impressively shown that all ML algorithms mentioned above gained performance when used in combination with deep neural networks.