.png){kind=icon}
.png){kind=icon}
Return
Data-driven bayesian-guided activation functions for multi-task pattern recognition
.png){kind=icon}
PRE
AI
DOI：10.1016/j.patcog.2025.112911.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• A synergistic optimisation neural network strategy of gradient descent and Bayesian inference is proposed to effectively alleviate the limitation that traditional methods are prone to fall into local optimums. • Data-driven prior distribution construction methods to overcome the artificial prior selection bias inherent in traditional Bayesian methods. • Generalised gated composite activation function dynamically adjusts basis function combinations to enhance nonlinear expressivity in multitask pattern recognition.