2025/09/09 by Zakeria Sharif Ali, Ali, Zakeria Sharif · 4 voices
Computer Science · #Artificial neural network #Discriminative model #Domain Adaptation and Few-Shot Learning #Gaussian #Gaussian Processes and Bayesian Inference #Generative Adversarial Networks and Image Synthesis #Mixture model #Pattern recognition (psychology) #Probabilistic logic #Scalability #Univariate
paper · pdf · doi:10.48550/arxiv.2509.07569
published in arXiv (Cornell University) (Cornell University)
openalex publication_date 2025/09/09 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/05
This paper introduces the Univariate Gaussian Mixture Model Neural Network (uGMM-NN), a novel neural architecture that embeds probabilistic reasoning directly into the computational units of deep networks. Unlike traditional neurons, which apply weighted sums followed by fixed non-linearities, each uGMM-NN node parameterizes its activations as a univariate Gaussian mixture, with learnable means, variances, and mixing coefficients. This design enables richer representations by capturing multimodality and uncertainty at the level of individual neurons, while retaining the scalability of standard feed-forward networks. We demonstrate that uGMM-NN can achieve competitive discriminative performance compared to conventional multilayer perceptrons, while additionally offering a probabilistic interpretation of activations. The proposed framework provides a foundation for integrating uncertainty-aware components into modern neural architectures, opening new directions for both discriminative and generative modeling.