![[2602.13864] Evolving Multi-Channel Confidence-Aware Activation Functions for Missing Data with Channel Propagation](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.13864] Evolving Multi-Channel Confidence-Aware Activation Functions for Missing Data with Channel Propagation
Summary
This paper presents a novel approach to activation functions in neural networks that incorporates missing data and confidence scores, enhancing model performance in scenarios with incomplete datasets.
Why It Matters
The research addresses a critical challenge in machine learning: the impact of missing data on model accuracy. By proposing new activation functions that account for missingness and confidence, this work aims to improve predictive performance, which is vital for applications in various fields where data incompleteness is common.
Key Takeaways
- Introduces Three-Channel Evolved Activations (3C-EA) for better handling of missing data.
- Presents ChannelProp, an algorithm for propagating missingness and confidence through neural networks.
- Demonstrates improved classification performance under various missing data scenarios.
Computer Science > Neural and Evolutionary Computing arXiv:2602.13864 (cs) [Submitted on 14 Feb 2026] Title:Evolving Multi-Channel Confidence-Aware Activation Functions for Missing Data with Channel Propagation Authors:Naeem Shahabi Sani, Ferial Najiantabriz, Shayan Shafaei, Dean F. Hougen View a PDF of the paper titled Evolving Multi-Channel Confidence-Aware Activation Functions for Missing Data with Channel Propagation, by Naeem Shahabi Sani and 3 other authors View PDF HTML (experimental) Abstract:Learning in the presence of missing data can result in biased predictions and poor generalizability, among other difficulties, which data imputation methods only partially address. In neural networks, activation functions significantly affect performance yet typical options (e.g., ReLU, Swish) operate only on feature values and do not account for missingness indicators or confidence scores. We propose Three-Channel Evolved Activations (3C-EA), which we evolve using Genetic Programming to produce multivariate activation functions f(x, m, c) in the form of trees that take (i) the feature value x, (ii) a missingness indicator m, and (iii) an imputation confidence score c. To make these activations useful beyond the input layer, we introduce ChannelProp, an algorithm that deterministically propagates missingness and confidence values via linear layers based on weight magnitudes, retaining reliability signals throughout the network. We evaluate 3C-EA and ChannelProp on datasets wit...
