REVIEW 23 cited by
Modular Duality in Deep Learning
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Modular Duality in Deep Learning
read the original abstract
An old idea in optimization theory says that since the gradient is a dual vector it may not be subtracted from the weights without first being mapped to the primal space where the weights reside. We take this idea seriously in this paper and construct such a duality map for general neural networks. Our map, which we call modular dualization, forms a unifying theoretical basis for training algorithms that are a) fast and b) scalable. Modular dualization involves first assigning operator norms to layers based on the semantics of each layer, and then using these layerwise norms to recursively induce a duality map on the weight space of the full neural architecture. We conclude by deriving GPU-friendly algorithms for dualizing Embed, Linear and Conv2D layers -- the latter two methods are based on a rectangular Newton-Schulz iteration (Kovarik, 1970; Bj\"orck & Bowie, 1971). A variant of our methods was used to set speed records for training NanoGPT. Overall, we hope that our theory of modular duality will yield a next generation of fast and scalable optimizers for general neural architectures.
Forward citations
Cited by 23 Pith papers
-
Dead-Direction Conditioners: Gauge-Equivariant Preconditioning for Deep Networks
Dead-Direction Conditioners provide gauge-equivariant preconditioning by conditioning optimizer state on symmetry orbits, yielding improved resistance to over-training collapse and higher detection of dead directions ...
-
Why Muon Outperforms Adam: A Curvature Perspective
Muon outperforms Adam by reducing curvature penalty via lower Normalized Directional Sharpness, as shown via Taylor approximation on LLM training and proven on stylized quadratic problems with heterogeneous curvature.
-
How to Scale Mixture-of-Experts: From muP to the Maximally Scale-Stable Parameterization
The authors derive a Maximally Scale-Stable Parameterization (MSSP) for MoE models that achieves robust learning-rate transfer and monotonic performance gains with scale across co-scaling regimes of width, experts, an...
-
Intrinsic Muon: Spectral Optimization on Riemannian Matrix Manifolds
Intrinsic Muon provides closed-form linear maximization oracles on multiple Riemannian matrix manifolds for unitarily invariant norms, with convergence rates depending only on manifold dimension or rank.
-
A unified convergence theory for adaptive first-order methods in the nonconvex case, including AdaNorm, full and diagonal AdaGrad, Shampoo and Muo
A unified stochastic convergence theory is developed for adaptive preconditioned first-order methods including AdaGrad variants, Shampoo, and Muon in nonconvex optimization.
-
Training Deep Learning Models with Norm-Constrained LMOs
Scion is a new stochastic LMO-based optimizer family that unifies existing methods, supports unconstrained problems, and delivers hyperparameter transferability plus speedups on nanoGPT training.
-
Reassessing Muon for Matrix Factorization
Muon's advantage over AdamW is problem-dependent: it loses or ties on plain low-rank factorization and completion but wins on nonnegative matrix factorization.
-
SOAP, Muon, and Beyond: Pushing LLM Pretraining Scales
SOAP and Muon, stabilized by per-step QR eigenbasis updates and KL-Shampoo covariance accumulation, beat AdamW on large-batch LLM pretraining up to 100M-token batches.
-
Stochastic convergence of parallel asynchronous adaptive first-order methods
Introduces a class of asynchronous adaptive first-order methods and establishes O(1/sqrt t) convergence (up to logs) for non-convex stochastic optimization under reasonable assumptions.
-
Muon Does Not Converge on Convex Lipschitz Functions
Muon does not converge on convex Lipschitz functions regardless of learning rate, while error feedback restores theoretical convergence but degrades performance on CIFAR-10 and nanoGPT tasks.
-
Muon-OGD: Muon-based Spectral Orthogonal Gradient Projection for LLM Continual Learning
Muon-OGD introduces a spectral-norm constrained orthogonal projection method solved via dual iterations and Newton-Schulz approximations to improve stability-plasticity trade-off in sequential LLM adaptation.
-
OrScale: Orthogonalised Optimization with Layer-Wise Trust-Ratio Scaling
OrScale adds a Frobenius-norm trust-ratio layer-wise scaler to Muon’s orthogonalized updates, with per-layer calibration for language models, yielding higher CIFAR-10 accuracy and better language-model pre-training lo...
-
Orth-Dion: Eliminating Geometric Mismatch in Distributed Low-Rank Spectral Optimization
Orth-Dion uses QR factorization on the right factor instead of column normalization to eliminate the geometric mismatch in low-rank approximations of spectral optimizers like Muon, achieving O(sqrt(L_r/T)) rate under ...
-
Pro-KLShampoo: Projected KL-Shampoo with Whitening Recovered by Orthogonalization
Pro-KLShampoo projects KL-Shampoo preconditioners to a spike-and-flat parametric form on an r-dimensional subspace and recovers the full algebraic preconditioner via orthogonalization, outperforming KL-Shampoo on GPT-...
-
GaussianGPT: Towards Autoregressive 3D Gaussian Scene Generation
A causal transformer with 3D RoPE generates vector-quantized 3D Gaussian latent grids autoregressively, enabling unconditional synthesis, completion, and open-ended outpainting of indoor scenes.
-
Decoupling Variance and Scale-Invariant Updates in Adaptive Gradient Descent for Unified Vector and Matrix Optimization
DeVA_S8 reweights Muon's matrix-sign update in the matrix's eigenbasis with a singular-value signal-to-noise ratio, reaching target LLM validation perplexity with ~6.6% fewer tokens than Muon.
-
Preconditioned Norms: A Unified Framework for Steepest Descent, Quasi-Newton and Adaptive Methods
Preconditioned matrix norms unify steepest descent, quasi-Newton, and adaptive optimizers, revealing SGD, Adam, Muon, KL-Shampoo, SOAP, and SPlus as special cases and enabling new methods MuAdam and MuAdam-SANIA that ...
-
Token Geometry
Ember matches Adam on embedding/LM-head tables with O(V+D) state by factoring second moments and dropping first-moment momentum, and token trajectories look nearly 1D.
-
Token Geometry
Ember is a memory-efficient optimizer for token embeddings that exploits distinct gradient geometry and models token trajectories as 1D rays.
-
Pion: A Spectrum-Preserving Optimizer via Orthogonal Equivalence Transformation
Pion is an optimizer that preserves the singular values of weight matrices in LLM training by applying orthogonal equivalence transformations.
-
Muon-OGD: Muon-based Spectral Orthogonal Gradient Projection for LLM Continual Learning
Muon-OGD integrates Muon-style spectral-norm geometry with orthogonal gradient constraints to improve the stability-plasticity trade-off during sequential LLM adaptation.
-
HTMuon: Improving Muon via Heavy-Tailed Spectral Correction
HTMuon modifies Muon to produce heavier-tailed updates and weight spectra via HT-SR theory, yielding up to 0.98 lower perplexity on LLaMA pretraining and serving as a plug-in for other Muon variants.
-
Lifelong In-Context Learning with Transformers Requires Parametric Forms of Attention
Argues that parametric attention forms are necessary for lifelong in-context learning in transformers to maintain constant memory footprint over arbitrary sequence lengths.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.