arXiv subjects

Joppe De Jonghe

Publications and source records attributed to Joppe De Jonghe.

5 recordsLinked to original sources

arXiv2026

Making Tensor Decoupling More Robust and Less Parametric

Tensor decoupling algorithms recover compact representations of multivariate functions by combining linear factor matrices with univariate internal nonlinearities. We identify two limitations of recent methods. First, their performance is not invariant to mismatched output or gradient magnitudes, and second, their spline internals require manual tuning of their degrees of freedom. We propose two lightweight fixes. The first is a preprocessing step that rescales the input Jacobian tensor, removing both the bias towards larger magnitude outputs and the ill-conditioning of certain factor updates caused by such a mismatch. The second replaces B-spline internals by penalized splines coupled with generalized cross-validation, making the method less dependent on the user selecting the right number of control points. Finally, we release an efficient open-source JAX implementation of our algorithm.

cs.LG↗

arXiv2026

Adaptive Subspace Modeling With Functional Tucker Decomposition

Tensors provide a structured representation for multidimensional data, yet discretization discards the underlying continuous structure when the data originate from continuous processes. We address this limitation by introducing a functional Tucker decomposition (FTD) that embeds a mode-wise continuity constraint directly into the factorization. The FTD models the continuous mode as a function in a reproducing kernel Hilbert space (RKHS), avoiding a prespecified basis while preserving the multilinear subspace structure of the Tucker model. We derive a reconstruction error bound for the continuous mode that quantifies the approximation quality when a subspace estimated on one domain is reused on another. This bound provides theoretical justification for subspace transfer, whose practical value we demonstrate on cross-domain classification tasks in hyperspectral imaging and multivariate time-series analysis.

stat.ML↗

arXiv2026

Robust Basis Spline Decoupling for the Compression of Transformer Models

Decoupling is a powerful modeling paradigm for representing multivariate functions as compositions of linear transformations and univariate nonlinear functions. A single-layer decoupling can be viewed as a fully connected neural network with a single hidden layer and flexible activation functions, providing a direct link with neural networks. Because of this, the use of decoupling methods has gained increasing attention in neural network domains, particularly compression, since it enables structured approximations with reduced parameter complexity. Existing tensor-based decoupling methods typically rely on polynomial or piecewise-linear parameterizations of the internal nonlinear functions, which can suffer from numerical instability or limited expressiveness. In this work, we introduce a B-spline-based decoupling framework that generalizes these existing approaches. By exploiting the local support and flexible smoothness control of B-splines, the proposed formulation yields a more numerically stable and expressive representation. We derive a constrained coupled matrix-tensor factorization and propose a robust alternating least-squares algorithm, called R-CMTF-BSD, incorporating normalization and Tikhonov regularization. The proposed method is validated through experiments on synthetic data and transformer model compression. Results on the Vision and Swin Transformer architectures demonstrate that B-spline decoupling enables substantial parameter reduction while maintaining competitive accuracy, making the R-CMTF-BSD algorithm a promising tool for structured neural network compression.

cs.LG↗

arXiv2026

Tensor-based Multi-layer Decoupling

The decoupling of multivariate functions is a powerful modeling paradigm for learning multivariate input-output relations from data. For the single-layer case, established CPD-based methods are available, but the multi-layer case remained largely unexplored. This work introduces a tensor-based framework for multi-layer decoupling, which is based on ParaTuck-type tensor decompositions and constrained optimization. We provide theoretical justification behind the considered tensor decompositions and parameterizations. Furthermore, we formulate a structured coupled matrix-tensor factorization that incorporates both Jacobian and function evaluations, together with a bilevel optimization approach for adaptively balancing first- and zeroth-order information. The feasibility of the proposed methodology is illustrated on synthetic systems, a nonlinear system identification benchmark and neural network compression.

eess.SY↗

arXiv2025

Non-parametric B-spline decoupling of multivariate functions

Many scientific fields and applications require compact representations of multivariate functions. For this problem, decoupling methods are powerful techniques for representing the multivariate functions as a combination of linear transformations and nonlinear univariate functions. This work introduces an efficient decoupling algorithm that leverages the use of B-splines to allow a non-parametric estimation of the decoupling's internal functions. The use of B-splines alleviates the problem of choosing an appropriate basis, as in parametric methods, but still allows an intuitive way to tweak the flexibility of the estimated functions. Besides the non-parametric property, the use of B-spline representations allows for easy integration of nonnegativity or monotonicity constraints on the function shapes, which is not possible for the currently available (non-)parametric decoupling methods. The proposed algorithm is illustrated on synthetic examples that highlight the flexibility of the B-spline representation and the ease with which a monotonicity constraint can be added. The examples also show that if monotonic functions are required, enforcing the constraint is necessary.

eess.SY↗