Latest Research

RPN 2: On Interdependence Function Learning (November 2024)

This paper proposes a redesign of the RPN architecture, introducing the new RPN 2 (i.e., Reconciled Polynomial Network version 2.0) model. As illustrate by Figure 1, RPN 2 incorporates a novel component, the interdependence functions, to explicitly model diverse relationships among both data instances and attributes. While we refer to this component as “interdependence”, this function actually captures a wide range of relationships within the input data, including structural interdependence, logical causality, statistical correlation, and numerical similarity or dissimilarity

RPN: Reconciled Polynomial Network (July 2024)

This paper proposes the task of "deep function learning" and introduce a novel deep function learning base model, i.e., the Reconciled Polynomial Network (RPN).
RPN has a versatile model architecture and attains superior modeling capabilities for diverse deep function learning tasks on various multi-modality datasets. RPN also provides a canonical representation for many existing machine learning and deep learning models, including but not limited to PGMs, kernel SVM, MLP and KAN.