Neural Networks and Their Internal Processes: From the ground up
Kutenkov, Ilia (2022)
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2022082319605
https://urn.fi/URN:NBN:fi:amk-2022082319605
Tiivistelmä
The aim of this thesis is to explain and practically show the operation of different types of neural networks. The thesis will prove that they all have the same primary idea at the core but have different goals, methods, and components. The thesis also shows a process of preparing a custom object detection dataset that could be used for training neural networks.
The author decided to show the neural networks from the ground up. First, he explains an abstract operation of a chosen tool. Then, a mathematical explanation is written. Finally, the code implementation or use case of a tool is shown.
As a result, the thesis has a lot of information about neural networks with associated code examples. It also explains processes that happen in neural networks during their forward and backward operations. Two different datasets were used to train neural networks. The thesis’ goals were successfully achieved.
The author decided to show the neural networks from the ground up. First, he explains an abstract operation of a chosen tool. Then, a mathematical explanation is written. Finally, the code implementation or use case of a tool is shown.
As a result, the thesis has a lot of information about neural networks with associated code examples. It also explains processes that happen in neural networks during their forward and backward operations. Two different datasets were used to train neural networks. The thesis’ goals were successfully achieved.