Saved in:
| Main Author: | |
|---|---|
| Format: | Recurso digital |
| Language: | |
| Published: |
Zenodo
2025
|
| Online Access: | https://doi.org/10.5281/zenodo.15766630 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866901649069441024 |
|---|---|
| author | Zhang, Jincheng |
| author_facet | Zhang, Jincheng |
| contents | <p><span>This paper innovatively incorporates the conflict theory ideas in sociology, especially the core views of Marx and Dahrendorf on social contradictions and class struggle, into the activation function design in deep learning. The specific approach is to map the contradictory tension of the opposing sides in the conflict theory, which both suppresses and interacts with each other, to the different response mechanisms of the activation function to the positive and negative parts of the input signal: the positive signal maintains the traditional incentive, while the negative signal introduces a complex expression that combines inhibition and nonlinear resistance, thereby simulating the dynamic interweaving tension relationship between the two sides of the social contradiction. By constructing this activation function based on the concept of conflict theory, we not only verified it in the multi-layer perceptron (MLP) model, but also emphasized the wide applicability and potential value of this activation mechanism in various neural network architectures. Experimental results show that the activation function based on conflict theory is superior to the traditional ReLU activation in multiple key performance indicators, reflecting the unique advantages of this theoretical method in improving the nonlinear expression and generalization ability of the model.</span></p> |
| format | Recurso digital |
| id | zenodo_https___doi_org_10_5281_zenodo_15766630 |
| institution | Zenodo |
| language | |
| publishDate | 2025 |
| publisher | Zenodo |
| record_format | zenodo |
| spellingShingle | A Breakthrough Activation Function Inspired by Conflict Theory for Superior Neural Modeling Zhang, Jincheng <p><span>This paper innovatively incorporates the conflict theory ideas in sociology, especially the core views of Marx and Dahrendorf on social contradictions and class struggle, into the activation function design in deep learning. The specific approach is to map the contradictory tension of the opposing sides in the conflict theory, which both suppresses and interacts with each other, to the different response mechanisms of the activation function to the positive and negative parts of the input signal: the positive signal maintains the traditional incentive, while the negative signal introduces a complex expression that combines inhibition and nonlinear resistance, thereby simulating the dynamic interweaving tension relationship between the two sides of the social contradiction. By constructing this activation function based on the concept of conflict theory, we not only verified it in the multi-layer perceptron (MLP) model, but also emphasized the wide applicability and potential value of this activation mechanism in various neural network architectures. Experimental results show that the activation function based on conflict theory is superior to the traditional ReLU activation in multiple key performance indicators, reflecting the unique advantages of this theoretical method in improving the nonlinear expression and generalization ability of the model.</span></p> |
| title | A Breakthrough Activation Function Inspired by Conflict Theory for Superior Neural Modeling |
| url | https://doi.org/10.5281/zenodo.15766630 |