paper-with-me

Papers

Momentum Stiefel Optimizer, with Applications to Suitably-Orthogonal Attention, and Optimal Transport

2022-05-27 · Lingkai Kong, Yuqing Wang, Molei Tao

The problem of optimization on Stiefel manifold, i.e., minimizing functions of (not necessarily square) matrices that satisfy orthogonality constraints, has been extensively studied. Yet, a new approach is proposed based on, for the first time, an interplay between thoughtfully designed continuous and discrete dynamics. It leads to a gradient-based optimizer with intrinsically added momentum. This method exactly preserves the manifold structure but does not require additional operation to keep momentum in the changing (co)tangent space, and thus has low computational cost and pleasant accuracy. Its generalization to adaptive learning rates is also demonstrated. Notable performances are observed in practical tasks. For instance, we found that placing orthogonal constraints on attention heads of trained-from-scratch Vision Transformer [Dosovitskiy et al. 2022] could markedly improve its performance, when our optimizer is used, and it is better that each head is made orthogonal within itself but not necessarily to other heads. This optimizer also makes the useful notion of Projection Robust Wasserstein Distance [Paty & Cuturi 2019; Lin et al. 2020] for high-dim. optimal transport even more effective.

📄 PDF Abstract BibTeX arXiv:2205.14173

Code (1)

konglk1203/variationalstiefeloptimizer 공식 구현 pytorch

Methods 이 논문이 사용한 방법론

Multi-Head Attention 설명 없음
Attention 설명 없음
Linear Layer A Linear Layer is a projection $\mathbf{XW + b}$.
Softmax The Softmax output function transforms a previous layer's output into a vector of probabilities. It is commonly used for multiclass classification. Given an input vector $x$…
Absolute Position Encodings Absolute Position Encodings are a type of position embeddings for [Transformer-based models] where positional encodings are…
Dropout Dropout is a regularization technique for neural networks that drops a unit (along with connections) at training time with a specified probability $p$ (a common value is…
Position-Wise Feed-Forward Layer 설명 없음
BPE Byte Pair Encoding, or BPE, is a subword segmentation algorithm that encodes rare and unknown words as sequences of subword units. The intuition is that various word…

Similar Papers 제목 키워드 기반

Riemannian Optimization for LoRA on the Stiefel Manifold

2025-08-25 · Juneyoung Park, Minjae Kang, Seongbae Lee, Haegang Lee 외 arxiv

While powerful, large language models (LLMs) present significant fine-tuning challenges due to their size. Parameter-efficient fine-tuning (PEFT) methods like LoRA provide solutions, yet suffer from critical optimizer in…

parameter-efficient fine-tuning

AuON: A Linear-time Alternative to Orthogonal Momentum Updates

2025-09-29 · Dipan Maity arxiv

Orthogonal momentum gradient updates have emerged to overcome the limitations of vector-based optimizers like Adam. The vector-based optimizer Adam suffers from high memory costs and ill-conditioned momentum gradient upd…

FedSPDnet: Geometry-Aware Federated Deep Learning with SPDnet

2026-04-24 · Thibault Pautrel, Florent Bouchard, Ammar Mian, Guillaume Ginolhac arxiv

We introduce two federated learning frameworks for the classical SPDnet model operating on symmetric positive definite (SPD) matrices with Stiefel-constrained parameters. Unlike standard Euclidean averaging, which violat…

Federated Learning

Generalizing Adam to Manifolds for Efficiently Training Transformers

2023-05-26 · Benedikt Brantner

One of the primary reasons behind the success of neural networks has been the emergence of an array of new, highly-successful optimizers, perhaps most importantly the Adam optimizer. It is widely used for training neural…

Physical Intuition

FISMO: Fisher-Structured Momentum-Orthogonalized Optimizer

2026-01-29 · Chenrui Xu, Wenjing Yan, Ying-Jun Angela Zhang arxiv

Training large-scale neural networks requires solving nonconvex optimization where the choice of optimizer fundamentally determines both convergence behavior and computational efficiency. While adaptive methods like Adam…

Computational EfficiencyImage Classification