Selected work

Representative Papers

Inventory of the 12 007 Low-Dimensional Pseudo-Boolean Landscapes Invariant to Rank, Translation, and Rotation

Apr 07, 2026

This work proposes a robust landscape equivalence relation that simultaneously incorporates invariance under permutation, translation, and rotation, enabling a systematic classification of all pseudo-Boolean optimization functions—including non-injective cases—in dimensions one through three. By combining combinatorial enumeration with exhaustive verification, the study constructs for the first time a complete set of 12,007 invariant landscape classes, substantially fewer than those obtained under permutation invariance alone. The analysis reveals that non-injective functions dominate landscape diversity and elucidates intricate relationships among neutrality, deception, and the performance of hill-climbing algorithms. These findings provide a foundational resource for benchmark design and theoretical investigations in discrete optimization.

0 citationsRead paper

A Latent Representation Learning Framework for Hyperspectral Image Emulation in Remote Sensing

Mar 23, 2026

This work proposes the first hyperspectral image simulation framework based on latent representation learning, addressing the high computational cost of traditional radiative transfer models and their difficulty in jointly modeling spatial-spectral information. By leveraging a pretrained variational autoencoder (VAE) and enabling interpolation in the latent space from physical parameters, the method supports efficient spectral- and spatial-spectral-level simulation through either one-step or two-step training strategies. Introducing latent generative modeling to hyperspectral simulation for the first time, the approach significantly outperforms conventional regression-based methods on both PROSAIL-simulated data and real Sentinel-3 OLCI imagery, achieving superior performance in reconstruction accuracy, spectral fidelity, and downstream biophysical parameter retrieval tasks.

0 citationsRead paper

Structured Multidimensional Representation Learning for Large Language Models

Mar 05, 2026

This work addresses the parameter redundancy in embedding dimensions of large language models, which incurs substantial computational and memory costs during scaling. The authors propose L-Transformer, the first architecture to enable differentiable spectral decomposition of the embedding space. By leveraging the third-order tensor L-product, token embeddings are reshaped into spectral tensor slices, and attention and feed-forward operations are performed in the transform domain. This yields a Tensor Transformer composed of p independent spectral sub-transformers, introducing an inductive bias at the frequency level that supports slice-dependent frequency scaling to enhance generalization while remaining compatible with standard training pipelines. Experiments show that on IMDB and AG News, the encoder achieves up to 75% parameter reduction (with p=4) while maintaining competitive accuracy, and fully recovers baseline performance at BERT-base width.

0 citationsRead paper

Multidimensional Task Learning: A Unified Tensor Framework for Computer Vision Tasks

Feb 26, 2026

This work proposes a multidimensional task learning (MTL) framework grounded in the generalized Einstein MLP (GE-MLP), which operates directly in tensor space via Einstein products, eliminating the need to flatten high-dimensional data and thereby preserving its intrinsic structure. In contrast to conventional computer vision approaches that rely on matrix-based modeling and inherently disrupt data geometry through vectorization, the proposed framework unifies diverse tasks—such as classification, segmentation, and detection—as distinct dimensional configurations of tensors. This formulation not only strictly subsumes the representational capacity of traditional matrix methods but also theoretically demonstrates that mainstream vision tasks are special cases of MTL. Consequently, the framework enables the natural construction of more complex spatiotemporal or cross-modal tasks within a coherent end-to-end learning paradigm.

0 citationsRead paper

A Computationally Efficient Multidimensional Vision Transformer

Feb 23, 2026

This work addresses the high computational and memory overhead of Vision Transformers in deployment by introducing, for the first time, the tensor cosine product (C-product) into the Vision Transformer architecture. Leveraging the multilinear structure inherent in images and the orthogonality of the cosine transform, the proposed method constructs an efficient attention mechanism and structured feature representation. It achieves competitive accuracy while reducing the number of parameters to 1/C of the original model, where C denotes the number of channels, thereby substantially lowering model complexity. By integrating multilinear algebra with orthogonal transforms, this approach makes notable contributions both theoretically and in terms of practical deployment efficiency.

0 citationsRead paper

Recent publications

Latest Papers

Inventory of the 12 007 Low-Dimensional Pseudo-Boolean Landscapes Invariant to Rank, Translation, and Rotation

Apr 07, 2026

This work proposes a robust landscape equivalence relation that simultaneously incorporates invariance under permutation, translation, and rotation, enabling a systematic classification of all pseudo-Boolean optimization functions—including non-injective cases—in dimensions one through three. By combining combinatorial enumeration with exhaustive verification, the study constructs for the first time a complete set of 12,007 invariant landscape classes, substantially fewer than those obtained under permutation invariance alone. The analysis reveals that non-injective functions dominate landscape diversity and elucidates intricate relationships among neutrality, deception, and the performance of hill-climbing algorithms. These findings provide a foundational resource for benchmark design and theoretical investigations in discrete optimization.

0 citationsRead paper

A Latent Representation Learning Framework for Hyperspectral Image Emulation in Remote Sensing

Mar 23, 2026

This work proposes the first hyperspectral image simulation framework based on latent representation learning, addressing the high computational cost of traditional radiative transfer models and their difficulty in jointly modeling spatial-spectral information. By leveraging a pretrained variational autoencoder (VAE) and enabling interpolation in the latent space from physical parameters, the method supports efficient spectral- and spatial-spectral-level simulation through either one-step or two-step training strategies. Introducing latent generative modeling to hyperspectral simulation for the first time, the approach significantly outperforms conventional regression-based methods on both PROSAIL-simulated data and real Sentinel-3 OLCI imagery, achieving superior performance in reconstruction accuracy, spectral fidelity, and downstream biophysical parameter retrieval tasks.

0 citationsRead paper

Structured Multidimensional Representation Learning for Large Language Models

Mar 05, 2026

This work addresses the parameter redundancy in embedding dimensions of large language models, which incurs substantial computational and memory costs during scaling. The authors propose L-Transformer, the first architecture to enable differentiable spectral decomposition of the embedding space. By leveraging the third-order tensor L-product, token embeddings are reshaped into spectral tensor slices, and attention and feed-forward operations are performed in the transform domain. This yields a Tensor Transformer composed of p independent spectral sub-transformers, introducing an inductive bias at the frequency level that supports slice-dependent frequency scaling to enhance generalization while remaining compatible with standard training pipelines. Experiments show that on IMDB and AG News, the encoder achieves up to 75% parameter reduction (with p=4) while maintaining competitive accuracy, and fully recovers baseline performance at BERT-base width.

0 citationsRead paper

Multidimensional Task Learning: A Unified Tensor Framework for Computer Vision Tasks

Feb 26, 2026

This work proposes a multidimensional task learning (MTL) framework grounded in the generalized Einstein MLP (GE-MLP), which operates directly in tensor space via Einstein products, eliminating the need to flatten high-dimensional data and thereby preserving its intrinsic structure. In contrast to conventional computer vision approaches that rely on matrix-based modeling and inherently disrupt data geometry through vectorization, the proposed framework unifies diverse tasks—such as classification, segmentation, and detection—as distinct dimensional configurations of tensors. This formulation not only strictly subsumes the representational capacity of traditional matrix methods but also theoretically demonstrates that mainstream vision tasks are special cases of MTL. Consequently, the framework enables the natural construction of more complex spatiotemporal or cross-modal tasks within a coherent end-to-end learning paradigm.

0 citationsRead paper

A Computationally Efficient Multidimensional Vision Transformer

Feb 23, 2026

This work addresses the high computational and memory overhead of Vision Transformers in deployment by introducing, for the first time, the tensor cosine product (C-product) into the Vision Transformer architecture. Leveraging the multilinear structure inherent in images and the orthogonality of the cosine transform, the proposed method constructs an efficient attention mechanism and structured feature representation. It achieves competitive accuracy while reducing the number of parameters to 1/C of the original model, where C denotes the number of channels, thereby substantially lowering model complexity. By integrating multilinear algebra with orthogonal transforms, this approach makes notable contributions both theoretically and in terms of practical deployment efficiency.

0 citationsRead paper