Implementation of a combined metric for the assessment of different platform/DNN experiments

Context

Deep Learning (DL) refers to a machine learning method of artificial intelligence in which neural networks approximate any functions and solve a wide variety of tasks. In most cases, they achieve higher prediction accuracy than humans. However, they are rarely as efficient as humans. In order to increase e.g. energy efficiency, special hardware accelerators are designed, which, however, often degrade the accuracy by applying compression methods. Therefore, a trade-off between accuracy and resources exists, which must be individually assessed for each application. To facilitate the co-design of models and accelerators, accuracy and resources should be considered during the development in a combined way.

Goals

This thesis aims the implementation of a combined metric for the assessment of different platform and deep neural network (DNN) experiments. Therefore, a literature review will be performed to identify suitable similar metrics. This metric will be applied to rank DNN accelerators. It further will be adapted to quantify the accuracy-resource trade-off and assess the co-design in different domains.

Requirements

• Programming experience in Python and C++
• Basic knowledge of neural networks
• Motivation and interest in solving technical problems independently