KOMPARASI FUNGSI AKTIVASI NEURAL NETWORK PADA DATA TIME SERIES
Abstract
Abstract— The sophistication and success of machine learning in solving problems in various fields of artificial intelligence cannot be separated from the neural networks that form the basis of its algorithms. Meanwhile, the essence of a neural network lies in its activation function. However because so many activation function which are merged lately, it’s needed to search for proper activation function according to the model and it’s dataset used. In this study, the activation functions commonly used in machine learning models will be tested, namely; ReLU, GELU and SELU, for time series data in the form of stock prices. These activation functions are implemented in python and use the TensorFlow library, as well as a model developed based on the Convolutional Neural Network (CNN). From the results of this implementation, the results obtained with the CNN model, that the GELU activation function for time series data has the smallest loss value
Downloads
References
Ahmed, G. S. (2023). SELU (Scaled Exponential Linear Unit) Activation Function. Www.Iq.Opengenus.Org. https://iq.opengenus.org/scaled-exponential-linear-unit/
Akil, I. (2023). Face Detection Pada Gambar Dengan Menggunakan OpenCV Haar Cascade. INTI NUSA MANDIRI, 17(2).
Brownlee, J. (2021). How to Choose an Activation Function for Deep Learning. In Machine Learning Mastery (pp. 1–26). https://machinelearningmastery.com/choose-an-activation-function-for-deep-learning/
Gustineli, M. (2022). A survey on recently proposed activation functions for Deep Learning. 1–7. http://arxiv.org/abs/2204.02921
Hackersrealm. (2021). Normalize data using Max Absolute & Min Max Scaling | Machine Learning | Python. Www.Hackersrealm.Net. https://www.hackersrealm.net/post/normalize-data-using-max-absolute-min-max-scaling
Lederer, J. (2021). Activation Functions in Artificial Neural Networks: A Systematic Overview. 1–42. http://arxiv.org/abs/2101.09957
Nwankpa, C., Ijomah, W., Gachagan, A., & Marshall, S. (2018). Activation Functions: Comparison of trends in Practice and Research for Deep Learning. 1–20. http://arxiv.org/abs/1811.03378
Pretorius, A. M., Barnard, E., & Davel, M. H. (2019). ReLU and sigmoidal activation functions. CEUR Workshop Proceedings, 2540, 37–48.
Raschka, S., Patterson, J., & Nolet, C. (2020). Machine learning in python: Main developments and technology trends in data science, machine learning, and artificial intelligence. Information (Switzerland), 11(4), 1–48. https://doi.org/10.3390/info11040193
Ryabtsev, A. (2022). 8 Reasons Why Python is Good for AI and ML. Www.Jangostars.Com. https://djangostars.com/blog/why-python-is-good-for-artificial-intelligence-and-machine-learning/#:~:text=Python for machine learning is,and quickly see the results.
Shaw, S. (2022). Activation Functions Compared With Experiments. Www.Wandb.Ai. https://wandb.ai/shweta/Activation Functions/reports/Activation-Functions-Compared-With-Experiments--VmlldzoxMDQwOTQ
Szandała, T. (2018). Review and Comparison of Commonly Used Activation Functions for Deep Neural Networks. Available: https://link.springer.com/chapter/10.1007/978-981-15-5495-7_11.
Ureña-Pliego, M., Martínez-Marín, R., González-Rodrigo, B., & Marchamalo-Sacristán, M. (2023). Automatic Building Height Estimation: Machine Learning Models for Urban Image Analysis. Applied Sciences (Switzerland), 13(8). https://doi.org/10.3390/app13085037
Varshney, M., & Singh, P. (2021). Optimizing nonlinear activation function for convolutional neural networks. SIViP, 15.
Abstract viewed = 223 times
PDF downloaded = 357 times
Copyright (c) 2023 Ibnu Akil
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
Penulis yang menerbitkan jurnal ini menyetujui ketentuan berikut:
1. Penulis memegang hak cipta dan memberikan hak jurnal mengenai publikasi pertama dengan karya yang dilisensikan secara bersamaan di bawah Creative Commons Attribution 4.0 International License. yang memungkinkan orang lain untuk berbagi karya dengan pengakuan atas karya penulis dan publikasi awal pada jurnal.
2. Penulis dapat memasukkan pengaturan kontrak tambahan yang terpisah untuk distribusi non-eksklusif dari versi jurnal yang diterbitkan (misalnya, mengirimkannya ke repositori institusional atau menerbitkannya dalam sebuah buku), dengan pengakuan atas publikasi awalnya pada Jurnal.
3. Penulis diizinkan dan didorong untuk memposting karya mereka secara online (misalnya, dalam penyimpanan institusional atau di situs web mereka) sebelum dan selama proses pengiriman, karena hal itu dapat menghasilkan pertukaran yang produktif, serta kutipan dari karya yang diterbitkan sebelumnya.
