| Title |
Assigning different activation functions in artificial neural networks with the goal of achieving higher prediction accuracy |
| Authors |
Baravykas, Gytis ; Kardoka, Justas ; Grigaliūnas, Domas ; Naujokaitis, Darius |
| DOI |
10.15388/Proceedings.2024.44 |
| Full Text |
|
| Is Part of |
IVUS2024: 29th international conference "Information society and university studies", Vilnius University, Kaunas Faculty, Kaunas, Lithuania, May 17th, 2024: abstracts.. Vilnius : Vilniaus universiteto leidykla. 2024, p. 38 |
| Abstract [eng] |
The research paper explores the concept of using multiple activation functions in artificial neural networks and investigates their impact on model performance. The experiments conducted on various models such as AlexNet, ResNet50, TuNet, and SimpleNN reveal insights into the effectiveness of different activation function combinations. The results indicate that using multiple activation functions can lead to modest improvements in model performance, particularly in image segmentation tasks where modifications to the UNet architecture show significant enhancements. However, for time series regression/forecasting tasks, the experiments demonstrate that using multiple activation functions does not significantly improve prediction accuracy. Therefore, the paper concludes that while there are some benefits to using multiple activation functions in certain scenarios, the choice of activation function should be based on the specific task and dataset. |
| Published |
Vilnius : Vilniaus universiteto leidykla |
| Type |
Conference paper |
| Language |
English |
| Publication date |
2024 |
| CC license |
|
