Muon in Vision Transformers: Optimizer-Recipe Interactions and Gradient Spectra

Muon is a recently developed matrix-aware optimizer that has shown strong results in transformer training, but its behavior in vision transformers (ViTs) is not yet well understood. We study Muon for ViT training, largely on ImageNet-100 and Pl@ntNet-300K, comparing against AdamW under standard vision recipes involving mixup, cutmix, smoothing, and random augmentation and erasing. Muon consistently outperforms AdamW, with especially large gains on long-tailed Pl@ntNet macro top-1. These gains are also recipe-dependent, where Muon benefits much more than AdamW from advanced and significant data augmentation techniques. To understand this interaction, we analyze the singular-value structure of matrix gradients throughout the ViT. Within Muon training runs, removing heavy data augmentation induces a late-training spectral concentration and mode collapse in gradient matrices, primarily in deep MLP-down blocks. Under a fixed "full" augmentation recipe, the clearest Muon-AdamW contrast appears instead in QKV gradients, where AdamW gradient energy remains concentrated in a much narrower basis while Muon spreads energy across substantially more singular modes. Muon in ViTs is therefore best understood as an optimizer-recipe interaction. Under a fixed recipe, Muon differs from AdamW most clearly in attention projections, where its gradients consist of a broader spectral basis. Within Muon, a full training recipe is important for preventing late spectral concentration and mode collapse in deep feedforward blocks. We further demonstrate efficacy in training ViTs on image segmentation and masked autoencoder models, where Muon outperforms AdamW in all settings considered.

Beyond Muon: MUD (MomentUm Decorrelation) for Faster Transformer Training

Orthogonalized-momentum optimizers such as Muon improve transformer training by approximately whitening/orthogonalizing matrix-valued momentum updates via a short polar-decomposition iteration. However, polar-factor approximations typically require multiple large matrix multiplications, and the resulting overhead can be substantial and hardware-dependent. We introduce MUD (MomentUm Decorrelation), a complementary whitening approach that replaces Muon's polar update with a triangular (Cholesky-like) whitening surrogate inspired by classical Gram--Schmidt and Gauss-Seidel ideas. We show that row-orthonormal matrices are fixed points of the MUD map, relate the inner step to symmetric Gauss-Seidel preconditioning of the Gram matrix, and prove quadratic local convergence near the fixed point. In terms of time-to-perplexity, MUD yields consistent 10-50\% wall-clock improvements over tuned AdamW and Muon in time-to-perplexity, typically converging slightly slower per step than Muon but with substantially lower optimizer overhead -- relative to Muon, MUD improves peak tokens/s by roughly $1.3-2.6\times$ across most settings and up to nearly $3\times$ on GPT-2 large on an A100. We also demonstrate training a ESM-2 150M protein language model, where MUD matches Muon-level validation perplexity in significantly less wall-clock time.

Multilevel Training for Kolmogorov Arnold Networks

Algorithmic speedup of training common neural architectures is made difficult by the lack of structure guaranteed by the function compositions inherent to such networks. In contrast to multilayer perceptrons (MLPs), Kolmogorov-Arnold networks (KANs) provide more structure by expanding learned activations in a specified basis. This paper exploits this structure to develop practical algorithms and theoretical insights, yielding training speedup via multilevel training for KANs. To do so, we first establish an equivalence between KANs with spline basis functions and multichannel MLPs with power ReLU activations through a linear change of basis. We then analyze how this change of basis affects the geometry of gradient-based optimization with respect to spline knots. The KANs change-of-basis motivates a multilevel training approach, where we train a sequence of KANs naturally defined through a uniform refinement of spline knots with analytic geometric interpolation operators between models. The interpolation scheme enables a ``properly nested hierarchy'' of architectures, ensuring that interpolation to a fine model preserves the progress made on coarse models, while the compact support of spline basis functions ensures complementary optimization on subsequent levels. Numerical experiments demonstrate that our multilevel training approach can achieve orders of magnitude improvement in accuracy over conventional methods to train comparable KANs or MLPs, particularly for physics informed neural networks. Finally, this work demonstrates how principled design of neural networks can lead to exploitable structure, and in this case, multilevel algorithms that can dramatically improve training performance.

Learning physical unknowns from hydrodynamic shock and material interface features in ICF capsule implosions

In high energy density physics (HEDP) and inertial confinement fusion (ICF), predictive modeling is complicated by uncertainty in parameters that characterize various aspects of the modeled system, such as those characterizing material properties, equation of state (EOS), opacities, and initial conditions. Typically, however, these parameters are not directly observable. What is observed instead is a time sequence of radiographic projections using X-rays. In this work, we define a set of sparse hydrodynamic features derived from the outgoing shock profile and outer material edge, which can be obtained from radiographic measurements, to directly infer such parameters. Our machine learning (ML)-based methodology involves a pipeline of two architectures, a radiograph-to-features network (R2FNet) and a features-to-parameters network (F2PNet), that are trained independently and later combined to approximate a posterior distribution for the parameters from radiographs. We show that the estimated parameters can be used in a hydrodynamics code to obtain density fields and hydrodynamic shock and outer edge features that are consistent with the data. Finally, we demonstrate that features resulting from an unknown EOS model can be successfully mapped onto parameters of a chosen analytical EOS model, implying that network predictions are learning physics, with a degree of invariance to the underlying choice of EOS model.