This repository includes different approaches to implement signal preprocessing techniques such as DPD making use of machine learning methods. The machine learning methods applied in this repository were manually implemented, in order to get and give better understanding of the applied solutions.
A brief and good explanation of the Random Forest technique can be read in https://towardsdatascience.com/understanding-random-forest-58381e0602d2. The concept was first included in https://www.stat.berkeley.edu/~breiman/randomforest2001.pdf. This method is a popular regression technique, but it can also be used as variable selector (see https://hal.archives-ouvertes.fr/hal-00755489/document). As a basic explanation, a Random Forest is formed by many decision trees (see https://medium.com/@chiragsehra42/decision-trees-explained-easily-28f23241248).
ANNs are tha basic deep learning tehcnique. It is the most popular AI method and others are just modifications of this one. A good insight can be found in https://www.sciencedirect.com/science/article/abs/pii/S0731708599002721. A Complex Valued Neural Network is just a implementation of ANNs with support of complex input and output numbers (see https://arxiv.org/abs/1905.12321).
The results of the application of these techniques to real power amplifiers (PA) behavior can be seen in https://www.researchgate.net/publication/339975273_Determining_a_Digital_Predistorter_Model_Structure_for_Wideband_Power_Amplifiers_through_Random_Forest and https://idus.us.es/bitstream/handle/11441/89549/TFG-2615-ALVAREZ.pdf?sequence=1&isAllowed=y. These documents give also context about previous and further line of investigation on this topic.