Table of Links Abstract and 1. Introduction Abstract and 1. Introduction Abstract and 1. Introduction Related Work


Low Rank Adaptation
3.1 LoRA and 3.2 Limitation of LoRA
3.3 ReLoRA*


Sparse Spectral Training
4.1 Preliminaries and 4.2 Gradient Update of U, VT with Σ
4.3 Why SVD Initialization is Important
4.4 SST Balances Exploitation and Exploration
4.5 Memory-Efficient Implementation for SST and 4.6 Sparsity of SST


Experiments
5.1 Machine Translation
5.2 Natural Language Generation
5.3 Hyperbolic Graph Neural Networks


Conclusion and Discussion


Broader Impacts and References Related Work Related Work Related Work Related Work Low Rank Adaptation
3.1 LoRA and 3.2 Limitation of LoRA
3.3 ReLoRA* Low Rank Adaptation Low Rank Adaptation 3.1 LoRA and 3.2 Limitation of LoRA 3.1 LoRA and 3.2 Limitation of LoRA 3.3 ReLoRA* 3.3 ReLoRA* Sparse Spectral Training
4.1 Preliminaries and 4.2 Gradient Update of U, VT with Σ
4.3 Why SVD Initialization is Important
4.4 SST Balances Exploitation and Exploration
4.5 Memory-Efficient Implementation for SST and 4.6 Sparsity of SST Sparse Spectral Training Sparse Spectral Training 4.1 Preliminaries and 4.2 Gradient Update of U, VT with Σ 4.1 Preliminaries and 4.2 Gradient Update of U, VT with Σ 4.3 Why SVD Initialization is Important 4.3 Why SVD Initialization is Important 4.4 SST Balances Exploitation and Exploration 4.4 SST Balances Exploitation and Exploration 4.5 Memory-Efficient Implementation for SST and 4.6 Sparsity of SST 4.5 Memory-Efficient Implementation for SST and 4.6 Sparsity of SST Experiments
5.1 Machine Translation
5.2 Natural Language Generation
5.3 Hyperbolic Graph Neural Networks Experiments Experiments Experiments 5.1 Machine Translation 5.1 Machine Translation 5.2 Natural Language Generation 5.2 Natural Language Generation 5.3 Hyperbolic Graph Neural Networks 5.3 Hyperbolic Graph Neural Networks Conclusion and Discussion Conclusion and Discussion Conclusion and Discussion Conclusion and Discussion Broader Impacts and References Broader Impacts and References Broader Impacts and References Broader Impacts and References Supplementary Information Supplementary Information A. Algorithm of Sparse Spectral Training A. Algorithm of Sparse Spectral Training B. Proof of Gradient of Sparse Spectral Layer B. Proof of Gradient of Sparse Spectral Layer C. Proof of Decomposition of Gradient of Weight C. Proof of Decomposition of Gradient of Weight D. Proof of Advantage of Enhanced Gradient over Default Gradient D. Proof of Advantage of Enhanced Gradient over Default Gradient E. Proof of Zero Distortion with SVD Initialization E. Proof of Zero Distortion with SVD Initialization F. Experiment Details F. Experiment Details G. Singular Value Pruning G. Singular Value Pruning H. Evaluating SST and GaLore: Complementary Approaches to Memory Efficiency H. Evaluating SST and GaLore: Complementary Approaches to Memory Efficiency I. Ablation Study I. Ablation Study 5 Experiments To validate our Sparse Spectral Training (SST) approach, we conducted experiments on both Euclidean and hyperbolic neural networks, demonstrating the generalization of SST across various neural network architectures and embedding geometries. We compared SST with full-rank training, LoRA, and ReLoRA*. The key distinctions between ReLoRA* and ReLoRA [5] is that ReLoRA includes a full-rank training as "warm start", making it not an end-to-end memory-efficient method. Moreover, ReLoRA* resets all optimizer states for low-rank parameters, unlike ReLoRA, which resets 99%. For our experiments, all linear layers in the baseline models were modified to their low-rank counterparts. Hyperparameters and implementation details are provided in Appendix F. Further comparisons of SST with the contemporaneous work GaLore [16] are elaborated in Appendix H, highlighting SST’s superior performance in low-rank configurations. Ablation studies are documented in Appendix I. 5.1 Machine Translation We employ the vanilla transformer [10] as the Euclidean transformer and HyboNet [12] as the hyperbolic transformer. Our experiments include three widely-used machine translation datasets: IWSLT’14 English-to-German [33], IWSLT’17 German-to-English [34], and Multi30K German-toEnglish [35]. For IWSLT’14, the hyperparameters are aligned with those from HyboNet. Table 1 presents BLEU scores for IWSLT’14 across various dimensions and ranks (r). The results confirm that SST consistently outperforms other low-rank methods. Notably, some BLEU scores for the hyperbolic transformer are zero, due to the training process encountering NaN losses, whereas SST maintains stability throughout. Previous hyperbolic neural network articles have predominantly focused on low-dimensional configurations [25, 36, 37]. A key characteristic of hyperbolic space is its exponential growth in volume with distance from a reference point, which is significantly more rapid than the polynomial growth seen in Euclidean space [38]. This expansive nature makes hyperbolic spaces particularly prone to overfitting as dimensionality increases. By imposing constraints on the parameter search space of hyperbolic neural networks, SST prevents the overfitting typically associated with such high-dimensional settings. This spectral sparse constraint enhances the stability and robustness of our models, ensuring consistent performance during training. Further comparative results on the Multi30K and IWSLT’17 datasets using the standard dimensions for vanilla Euclidean transformers are documented in Table 2. Here, SST not only surpasses other low-rank methods but also demonstrates superior performance compared to full-rank training. Authors:
(1) Jialin Zhao, Center for Complex Network Intelligence (CCNI), Tsinghua Laboratory of Brain and Intelligence (THBI) and Department of Computer Science;
(2) Yingtao Zhang, Center for Complex Network Intelligence (CCNI), Tsinghua Laboratory of Brain and Intelligence (THBI) and Department of Computer Science;
(3) Xinghang Li, Department of Computer Science;
(4) Huaping Liu, Department of Computer Science;
(5) Carlo Vittorio Cannistraci, Center for Complex Network Intelligence (CCNI), Tsinghua Laboratory of Brain and Intelligence (THBI), Department of Computer Science, and Department of Biomedical Engineering Tsinghua University, Beijing, China. Authors: Authors: (1) Jialin Zhao, Center for Complex Network Intelligence (CCNI), Tsinghua Laboratory of Brain and Intelligence (THBI) and Department of Computer Science; (2) Yingtao Zhang, Center for Complex Network Intelligence (CCNI), Tsinghua Laboratory of Brain and Intelligence (THBI) and Department of Computer Science; (3) Xinghang Li, Department of Computer Science; (4) Huaping Liu, Department of Computer Science; (5) Carlo Vittorio Cannistraci, Center for Complex Network Intelligence (CCNI), Tsinghua Laboratory of Brain and Intelligence (THBI), Department of Computer Science, and Department of Biomedical Engineering Tsinghua University, Beijing, China. This paper is available on arxiv under CC by 4.0 Deed (Attribution 4.0 International) license. This paper is available on arxiv under CC by 4.0 Deed (Attribution 4.0 International) license. available on arxiv available on arxiv

Part of HackerNoon's growing list of open-source research papers, promoting free access to academic material.

Generalizing Sparse Spectral Training Across Euclidean and Hyperbolic Architectures

About Author

Comments

TOPICS

THIS ARTICLE WAS FEATURED IN

Related Stories

Untitled Story

A Comparative Performance Analysis of SymTax on Five Citation Recommendation Datasets

New Training Method Cuts Neural Network Memory Costs Without Losing Accuracy

Breaking Down Low-Rank Adaptation and Its Next Evolution, ReLoRA

Why Sparse Spectral Training Might Replace LoRA in AI Model Optimization

The Geometric Revolution That's Making Computer Vision More Efficient

A Comparative Performance Analysis of SymTax on Five Citation Recommendation Datasets

New Training Method Cuts Neural Network Memory Costs Without Losing Accuracy

Breaking Down Low-Rank Adaptation and Its Next Evolution, ReLoRA

Why Sparse Spectral Training Might Replace LoRA in AI Model Optimization

The Geometric Revolution That's Making Computer Vision More Efficient

Light-Mode

Classic

Newspaper

Dark-Mode

Neon Noir

Minty

HN StartUps