PALIM

Goal / description

The aim of the PALIM research project is therefore to systematically investigate the suitability of modern deep learning methods (transformers, TCN) and performance optimizations (binary neural networks, mixed precision) for high-dimensional (mostly binary) time series in comparison to classical time series algorithms such as HMM or FSM using the example of a highly automated production line in the automotive industry and to find ideal trade-offs between the models or through model combinations.