论文检索

Deep Learning · Theory

Alexandru Crăciun, Debarghya Ghoshdastidar

The theory of training deep networks has become a central question of modern machine learning and has inspired many practical advancements. In particular, the gradient descent (GD) optimization algorithm has been extensively studied in recent years. A key assumption about GD has appeared in several recent works: the \emph{GD map is non-singular} --- it preserves sets of measure zero under preimages. Crucially, this assumption has been used to prove that GD avoids saddle points and maxima, and to establish the existence of a computable quantity that determines the convergence to global minima (both for GD and stochastic GD). However, the current literature either assumes the non-singularity of the GD map or imposes restrictive assumptions, such as Lipschitz smoothness of the loss (for example, Lipschitzness does not hold for deep ReLU networks with the cross-entropy loss) and restricts the analysis to GD with small step-sizes. In this paper, we investigate the neural network map as a function on the space of weights and biases. We also prove, for the first time, the non-singularity of the gradient descent (GD) map on the loss landscape of realistic neural network architectures (with fully connected, convolutional, or softmax attention layers) and piecewise analytic activations (which includes sigmoid, ReLU, leaky ReLU, etc.) for almost all step-sizes. Our work significantly extends the existing results on the convergence of GD and SGD by guaranteeing that they apply to practical neural network settings and has the potential to unlock further exploration of learning dynamics.

Optimization · Discrete Optimization

Nikolaos Karalias, Akbar Rafiey, Yifei Xu, Zhishang Luo, Behrooz Tahmasebi, Connie Jiang, Stefanie Jegelka

Self-Supervised Learning (SSL) for Combinatorial Optimization (CO) is an emerging paradigm for solving combinatorial problems using neural networks. In this paper, we address a central challenge of SSL for CO: solving problems with discrete constraints. We design an end-to-end differentiable framework that enables us to solve discrete constrained optimization problems with neural networks. Concretely, we leverage algorithmic techniques from the literature on convex geometry and Carathéodory's theorem to decompose neural network outputs into convex combinations of polytope corners that correspond to feasible sets. This decomposition-based approach enables self-supervised training but also ensures efficient quality-preserving rounding of the neural net output into feasible solutions. Extensive experiments in cardinality-constrained optimization show that our approach can consistently outperform neural baselines. We further provide worked-out examples of how our method can be applied beyond cardinality-constrained problems to a diverse set of combinatorial optimization tasks, including finding independent sets in graphs, and solving matroid-constrained problems.

Optimization

Connor Magoon, Fengyu Yang, Noam Aigerman, Shahar Kovalsky

Differentiable optimization has attracted significant research interest, particularly for quadratic programming (QP). Existing approaches for differentiating the solution of a QP with respect to its defining parameters often rely on specific integrated solvers. This integration limits their applicability, including their use in neural network architectures and bi-level optimization tasks, restricting users to a narrow selection of solver choices. To address this limitation, we introduce **dQP**, a modular and solver-agnostic framework for plug-and-play differentiation of virtually any QP solver. A key insight we leverage to achieve modularity is that, once the active set of inequality constraints is known, both the solution and its derivative can be expressed using simplified linear systems that share the same matrix. This formulation fully decouples the computation of the QP solution from its differentiation. Building on this result, we provide a minimal-overhead, open-source implementation () that seamlessly integrates with over 15 state-of-the-art solvers. Comprehensive benchmark experiments demonstrate dQP’s robustness and scalability, particularly highlighting its advantages in large-scale sparse problems.

Optimization · Neural Network Optimization

Junno Yun, Yasar Utku Alcalar, Mehmet Akcakaya

Algorithm unrolling methods have proven powerful for solving the regularized least squares problem in computational magnetic resonance imaging (MRI). These approaches unfold an iterative algorithm with a fixed number of iterations, typically alternating between a neural network-based proximal operator for regularization, a data fidelity operation and auxiliary updates with learnable parameters. While the connection to optimization methods dictate that the proximal operator network should be shared across unrolls, this can introduce artifacts or blurring. Heuristically, practitioners have shown that using distinct networks may be beneficial, but this significantly increases the number of learnable parameters, making it challenging to prevent overfitting. To address these shortcomings, by taking inspirations from proximal operators with varying thresholds in approximate message passing (AMP) and the success of time-embedding in diffusion models, we propose a time-embedded algorithm unrolling scheme for inverse problems. Specifically, we introduce a novel perspective on the iteration-dependent proximal operation in vector AMP (VAMP) and the subsequent Onsager correction in the context of algorithm unrolling, framing them as a time-embedded neural network. Similarly, the scalar weights in the data fidelity operation and its associated Onsager correction are cast as time-dependent learnable parameters. Our extensive experiments on the fastMRI dataset, spanning various acceleration rates and datasets, demonstrate that our method effectively reduces aliasing artifacts and mitigates noise amplification, achieving state-of-the-art performance. Furthermore, we show that our time-embedding strategy extends to existing algorithm unrolling approaches, enhancing reconstruction quality without increasing the computational complexity significantly. Code available at https://github.com/JN-Yun/TE-Unrolling-MRI.

Optimization · Neural Network Optimization

Yehonathan Refael, Guy Smorodinsky, Tom Tirer, Ofir Lindenbaum

Low-rank gradient-based optimization methods have significantly improved memory efficiency during the training of large language models (LLMs), enabling operations within constrained hardware without sacrificing performance. However, these methods primarily emphasize memory savings, often overlooking potential acceleration in convergence due to their reliance on standard isotropic steepest descent techniques, which can perform suboptimally in the highly anisotropic landscapes typical of deep networks, particularly LLMs. In this paper, we propose SUMO (Subspace-Aware Moment-Orthogonalization), an optimizer that employs exact singular value decomposition (SVD) for moment orthogonalization within a dynamically adapted low-dimensional subspace, enabling norm-inducing steepest descent optimization steps. By explicitly aligning optimization steps with the spectral characteristics of the loss landscape, SUMO effectively mitigates approximation errors associated with commonly used methods like Newton-Schulz orthogonalization approximation. We theoretically establish an upper bound on these approximation errors, proving their dependence on the condition numbers of moments, conditions we analytically demonstrate are encountered during LLM training. Furthermore, we both theoretically and empirically illustrate that exact orthogonalization via SVD substantially improves convergence rates while reducing overall complexity. Empirical evaluations confirm that SUMO accelerates convergence, enhances stability, improves performance, and reduces memory requirements by up to 20\% compared to state-of-the-art methods.

Theory · Learning Theory

Han Bao, Shinsaku Sakaue, Yuki Takezawa

The gradient descent (GD) has been one of the most common optimizer in machine learning. In particular, the loss landscape of a neural network is typically sharpened during the initial phase of training, making the training dynamics hover on the edge of stability. This is beyond our standard understanding of GD convergence in the stable regime where arbitrarily chosen stepsize is sufficiently smaller than the edge of stability. Recently, Wu et al. (COLT2024) have showed that GD converges with arbitrary stepsize under linearly separable logistic regression. Although their analysis hinges on the self-bounding property of the logistic loss, which seems to be a cornerstone to establish a modified descent lemma, our pilot study shows that other loss functions without the self-bounding property can make GD converge with arbitrary stepsize. To further understand what property of a loss function matters in GD, we aim to show arbitrary-stepsize GD convergence for a general loss function based on the framework of \emph{Fenchel--Young losses}. We essentially leverage the classical perceptron argument to derive the convergence rate for achieving $\epsilon$-optimal loss, which is possible for a majority of Fenchel--Young losses. Among typical loss functions, the Tsallis entropy achieves the GD convergence rate $T=\Omega(\epsilon^{-1/2})$, and the R{\'e}nyi entropy achieves the far better rate $T=\Omega(\epsilon^{-1/3})$. We argue that these better rate is possible because of \emph{separation margin} of loss functions, instead of the self-bounding property.

Optimization · Neural Network Optimization

Antonio Orvieto, Robert Gower

Understanding the remarkable efficacy of Adam when training transformer-based language models has become a central research topic within the optimization community. To gain deeper insights, several simplifications of Adam have been proposed, such as the signed gradient and signed momentum methods. In this work, we conduct an extensive empirical study — training over 1,500 language models across different data configurations and scales — comparing Adam to several known simplified variants. We find that signed momentum methods are faster than SGD, but consistently underperform relative to Adam, even after careful tuning of momentum, clipping setting and learning rates. However, our analysis reveals a compelling option that preserves near-optimal performance while allowing for new insightful reformulations: constraining the Adam momentum parameters to be equal, $\beta_1=\beta_2$. Beyond robust performance, this choice affords new theoretical insights, highlights the "secret sauce" on top of signed momentum, and grants a precise statistical interpretation: we show that Adam in this setting implements a natural online algorithm for estimating the mean and variance of gradients—one that arises from a mean-field Gaussian variational inference perspective.

Deep Learning · Graph Neural Networks

Yixiao Zhou, Xiaoqing Lyu, Hongxiang Lin, Huiying Hu, Tuo Wang

Link sign prediction in signed bipartite graphs, which are extensively utilized across diverse domains such as social networks and recommendation systems, has recently emerged as a pivotal challenge. However, significant space and time complexities associated with the scalability of bipartite graphs pose substantial challenges, particularly in large-scale environments. To address these issues, this paper introduces the SignFlow Bipartite Subgraph Network (SBSN), balancing sublinear training memory growth through a heuristic subgraph extraction method integrated with a novel message passing module, with optimal inference efficiency achieved via the node feature distillation module. Our subgraph sampling approach reduces the graph size by focusing on neighborhoods around target links and employs an optimized directed message passing mechanism to aggregate critical structural patterns. This mechanism allows SBSN to efficiently learn rich local structural patterns essential for accurate sign prediction. Furthermore, to overcome the inefficiency of subgraph sampling-based models during inference, SBSN incorporates a node feature distillation module after the first training stage. This module distills subgraph features into node features, enabling fast inference while retaining the rich structural information of subgraphs. Experiments reveal that SBSN shows superior performance in both medium- and large-scale datasets, efficiently managing memory and computational resources, making it a scalable solution for extensive applications.

Connor Dunlop, Matthew Zheng, Kavana Venkatesh, Pinar Yanardag

Text-to-image (T2I) diffusion models have made remarkable strides in generating and editing high-fidelity images from text. Yet, these models remain fundamentally generic, failing to adapt to the nuanced aesthetic preferences of individual users. In this work, we present the first framework for personalized image editing in diffusion models, introducing Collaborative Direct Preference Optimization (C-DPO), a novel method that aligns image edits with user-specific preferences while leveraging collaborative signals from like-minded individuals. Our approach encodes each user as a node in a dynamic preference graph and learns embeddings via a lightweight graph neural network, enabling information sharing across users with overlapping visual tastes. We enhance a diffusion model's editing capabilities by integrating these personalized embeddings into a novel DPO objective, which jointly optimizes for individual alignment and neighborhood coherence. Comprehensive experiments, including user studies and quantitative benchmarks, demonstrate that our method consistently outperforms baselines in generating edits that are aligned with user preferences.

Applications · Chemistry and Drug Discovery

Jongha (Jon) Ryu, Samuel Zhou, Gregory Wornell

Spectral decomposition of linear operators plays a central role in many areas of machine learning and scientific computing. Recent work has explored training neural networks to approximate eigenfunctions of such operators, enabling scalable approaches to representation learning, dynamical systems, and partial differential equations (PDEs). In this paper, we revisit a classical optimization framework from the computational physics literature known as the *orbital minimization method* (OMM), originally proposed in the 1990s for solving eigenvalue problems in computational chemistry. We provide a simple linear-algebraic proof of the consistency of the OMM objective, and reveal connections between this method and several ideas that have appeared independently across different domains. Our primary goal is to justify its broader applicability in modern learning pipelines. We adapt this framework to train neural networks to decompose positive semidefinite operators, and demonstrate its practical advantages across a range of benchmark tasks. Our results highlight how revisiting classical numerical methods through the lens of modern theory and computation can provide not only a principled approach for deploying neural networks in numerical simulation, but also effective and scalable tools for machine learning.

Deep Learning · Graph Neural Networks

Yeyu Yan, Shuai Zheng, Wenjun Hui, Xiangkai Zhu, Chen Dong, Zhenfeng Zhu, Yao Zhao, Kunlun He

Graph condensation (GC) aims to distill the original graph into a small-scale graph, mitigating redundancy and accelerating GNN training. However, conventional GC approaches heavily rely on rigid GNNs and task-specific supervision. Such a dependency severely restricts their reusability and generalization across various tasks and architectures. In this work, we revisit the goal of ideal GC from the perspective of GNN optimization consistency, and then a generalized GC optimization objective is derived, by which those traditional GC methods can be viewed nicely as special cases of this optimization paradigm. Based on this, \textbf{Pre}-trained \textbf{G}raph \textbf{C}ondensation (\textbf{PreGC}) via optimal transport is proposed to transcend the limitations of task- and architecture-dependent GC methods. Specifically, a hybrid-interval graph diffusion augmentation is presented to suppress the weak generalization ability of the condensed graph on particular architectures by enhancing the uncertainty of node states. Meanwhile, the matching between optimal graph transport plan and representation transport plan is tactfully established to maintain semantic consistencies across source graph and condensed graph spaces, thereby freeing graph condensation from task dependencies. To further facilitate the adaptation of condensed graphs to various downstream tasks, a traceable semantic harmonizer from source nodes to condensed nodes is proposed to bridge semantic associations through the optimized representation transport plan in pre-training. Extensive experiments verify the superiority and versatility of PreGC, demonstrating its task-independent nature and seamless compatibility with arbitrary GNNs.

Optimization · Neural Network Optimization

Yi Fan, Yu-Bin Yang

The Mamba-type neural networks have gained significant popularity recently. To effectively and efficiently establish model architectures of Mamba, it is natural to introduce Neural Architecture Search (NAS) methods into Mamba. However, existing NAS methods tailored for Mamba are training-based, leading to substantial time and computational resource expenditure. To address this issue, and considering that Mamba2 is an improved version of the original Mamba, we propose a training-free NAS method specifically designed for Mamba2. Based on rank collapse in stacked State Space Duality (SSD) blocks, we design a proxy that only requires the computation of the transformation matrix and its gradient between two tensors within the network. Additionally, we develop a corresponding search space and introduce a novel approach for determining adjustable hyperparameter ranges. Experimental results show that our method outperforms all existing training-free NAS approaches in terms of both ranking correlation and the performance of search results for Mamba2 architecture. To the best of our knowledge, this is the first training-free NAS method designed for Mamba-type architectures. Our codes are available at https://github.com/fanyi-plus/tf-nas.

Deep Learning · Theory

Peter Súkeník, Christoph Lampert, Marco Mondelli

The empirical emergence of neural collapse---a surprising symmetry in the feature representations of the training data in the penultimate layer of deep neural networks---has spurred a line of theoretical research aimed at its understanding. However, existing work focuses on data-agnostic models or, when data structure is taken into account, it remains limited to multi-layer perceptrons. Our paper fills both these gaps by analyzing modern architectures in data-aware regime: we prove that global optima of deep regularized transformers and residual networks (ResNets) with LayerNorm trained with cross entropy or mean squared error loss are approximately collapsed, and the approximation gets tighter as the depth grows. More generally, we formally reduce any end-to-end large-depth ResNet or transformer training into an equivalent unconstrained features model, thus justifying its wide use in the literature even beyond data-agnostic settings. Our theoretical results are supported by experiments on computer vision and language datasets showing that, as the depth grows, neural collapse indeed becomes more prominent.

Deep Learning · Theory

Zukang Xu, Xing Hu, Qiang Wu, Dawei Yang

Large language models (LLMs) have demonstrated remarkable performance across a wide range of natural language processing tasks. However, their exponentially increasing parameters pose significant challenges for deployment on resource-constrained devices. Vector Quantization (VQ) shows great promise for low-bit quantization (e.g., 2 to 4 bits), but existing work faces two key challenges: unconstrained direction error and suboptimal bit allocation. In this paper, we propose RSAVQ, a novel VQ framework to enhance extremely low-bit quantization for LLMs. RSAVQ introduces two geometry-driven innovations that effectively mitigate above limitations: (1) Error Direction Sensitivity Guidance (EDSG), which leverages the Fisher information matrix (FIM)-induced Riemannian metric to project quantization errors onto low-sensitivity directions in the parameter space. Specifically, this projection is performed along the negative natural gradient direction, which effectively suppresses error expansion. (2) Weight Channel Sensitivity Guidance (WCSG) , which constructs a channel-wise sensitivity metric via FIM curvature analysis to dynamically guide bit resource allocation. The approach facilitates a globally optimal quantization solution within prescribed bit constraints. Experiments demonstrate that RSAVQ outperforms existing methods for LLMs. For example, in 2-bit quantization of LLaMA-3 8B, RSAVQ leads baselines like VPTQ and QuIP\# by 0.4 in perplexity (PPL) and 1.5 in zero-shot accuracy. This work offers a practical solution for constrained environments and a theoretical bridge between information geometry and the quantization of neural networks, advancing efficient deep learning.

Optimization · Neural Network Optimization

Jiwon Song, Dongwon Jo, Yulhwa Kim, jae-joon kim

Recent reasoning-focused language models achieve high accuracy by generating lengthy intermediate reasoning paths before producing final answers. While this approach is effective in solving problems that require logical thinking, long reasoning paths significantly increase memory usage and reduce throughput of token generation, limiting the practical deployment of such models. We propose Reasoning Path Compression (RPC), a training-free method that accelerates inference by leveraging the semantic sparsity of reasoning paths. RPC periodically compresses the KV cache by retaining cache entries that receive high importance score, which are computed using a selector window composed of recently generated queries. Experiments show that RPC improves generation throughput of QwQ-32B by up to 1.60$\times$ compared to the inference with full KV cache, with an accuracy drop of 1.2\% on the AIME 2024 benchmark. Our findings demonstrate that semantic sparsity in reasoning traces can be effectively exploited for compression, offering a practical path toward efficient deployment of reasoning LLMs. Our code is available at https://github.com/jiwonsong-dev/ReasoningPathCompression.

Optimization

Alexander Bodard, Panagiotis Patrinos

We study generalized smoothness in nonconvex optimization, focusing on $(L_0, L_1)$-smoothness and anisotropic smoothness. The former was empirically derived from practical neural network training examples, while the latter arises naturally in the analysis of nonlinearly preconditioned gradient methods. We introduce a new sufficient condition that encompasses both notions, reveals their close connection, and holds in key applications such as phase retrieval and matrix factorization. Leveraging tools from dynamical systems theory, we then show that nonlinear preconditioning—including gradient clipping—preserves the saddle point avoidance property of classical gradient descent. Crucially, the assumptions required for this analysis are actually satisfied in these applications, unlike in classical results that rely on restrictive Lipschitz smoothness conditions. We further analyze a perturbed variant that efficiently attains second-order stationarity with only logarithmic dependence on dimension, matching similar guarantees of classical gradient methods.

Theory · Game Theory

Deep Patel, Emmanouil-Vasileios Vlatakis-Gkaragkounis

Many emerging applications—such as adversarial training, AI alignment, and robust optimization—can be framed as zero-sum games between neural nets, with von Neumann–Nash equilibria (NE) capturing the desirable system behavior. While such games often involve non-convex non-concave objectives, empirical evidence shows that simple gradient methods frequently converge, suggesting a hidden geometric structure. In this paper, we provide a theoretical framework that explains this phenomenon through the lens of \emph{hidden convexity} and \emph{overparameterization}. We identify sufficient conditions spanning initialization, training dynamics, and network width—that guarantee global convergence to a NE in a broad class of non-convex min-max games. To our knowledge, this is the first such result for games that involve two-layer neural networks. Technically, our approach is twofold: (a) we derive a novel path-length bound for alternating gradient-descent-ascent scheme in min-max games; and (b) we show that games with hidden convex–concave geometry reduce to settings satisfying two-sided Polyak–Łojasiewicz (PL) and smoothness conditions, which hold with high probability under overparameterization, using tools from random matrix theory.

Applications · Neuroscience, Cognitive Science

Basile Confavreux, William Dorrell, Nishil Patel, Andrew Saxe

Synaptic plasticity is widely considered to be crucial to the brain’s ability to learn throughout life. Decades of theoretical work have therefore been invested in deriving and designing biologically plausible learning rules capable of granting various memory abilities to neural networks. Most of these theoretical approaches optimize directly for a desired memory function; but this procedure can lead to complex, finely-tuned rules, rendering them brittle to perturbations and difficult to implement in practice. Instead, we build on recent work that automatically discovers large numbers of candidate plasticity rules operating in recurrent spiking neural networks. Surprisingly, despite the fact that these rules are selected solely to achieve network stabilization, we observe across a range of network models -feedforward, recurrent; rate and spiking- that almost all these rules endow the network with simple forms of memory such as familiarity detection - seemingly by accident. To understand this phenomenon, we study an analytic toy model. We observe that memory arises from the degeneracy of weight matrices that stabilize a network: where the network lands in this space of stable weights depends on its past inputs---that is, memory. Even simple Hebbian plasticity rules can utilize this degeneracy, creating a zoo of memory abilities with various lifetimes. In practice, the larger the network and the more co-active plasticity rules in the system, the stronger the memory-by-accident phenomenon becomes. Overall, our findings suggest that activity-silent memory is a near-unavoidable consequence of stabilization. Simple forms of memory, such as familiarity or novelty detection, appear to be widely available resources for plastic brain networks, suggesting that they could form the raw materials that were later sculpted into higher-order cognitive abilities.

Applications · Neuroscience, Cognitive Science

Kabir Dabholkar, Omri Barak

Many natural systems, including neural circuits involved in decision making, are modeled as high-dimensional dynamical systems with multiple stable states. While existing analytical tools primarily describe behavior near stable equilibria, characterizing separatrices -- the manifolds that delineate boundaries between different basins of attraction -- remains challenging, particularly in high-dimensional settings. Here, we introduce a numerical framework leveraging Koopman Theory combined with Deep Neural Networks to effectively characterize separatrices. Specifically, we approximate Koopman Eigenfunctions (KEFs) associated with real positive eigenvalues, which vanish precisely at the separatrices. Utilizing these scalar KEFs, optimization methods efficiently locate separatrices even in complex systems. We demonstrate our approach on synthetic benchmarks, ecological network models, and high-dimensional recurrent neural networks trained on either neuroscience-inspired tasks or fit to real neural data. Moreover, we illustrate the practical utility of our method by designing optimal perturbations that can shift systems across separatrices, enabling predictions relevant to optogenetic stimulation experiments in neuroscience.

Social Aspects · Accountability, Transparency and Interpretability

Daniel Wurgaft, Ekdeep S Lubana, Core Francisco Park, Hidenori Tanaka, Gautam Reddy, Noah Goodman

Recent work analyzing in-context learning (ICL) has identified a broad set of strategies that describe model behavior in different experimental conditions. We aim to unify these findings by asking why a model learns these disparate strategies in the first place. Specifically, we start with the observation that when trained to learn a mixture of tasks, as is popular in the literature, the strategies learned by a model for performing ICL can be captured by a family of Bayesian predictors: a memorizing predictor, which assumes a discrete prior on the set of seen tasks, and a generalizing predictor, where the prior matches the underlying task distribution. Adopting the normative lens of rational analysis, where a learner’s behavior is explained as an optimal adaptation to data given computational constraints, we develop a hierarchical Bayesian framework that almost perfectly predicts Transformer next- token predictions throughout training—without assuming access to its weights. Under this framework, pretraining is viewed as a process of updating the posterior probability of different strategies, and inference-time behavior as a posterior- weighted average over these strategies’ predictions. Our framework draws on common assumptions about neural network learning dynamics, which make explicit a tradeoff between loss and complexity among candidate strategies: beyond how well it explains the data, a model’s preference towards implementing a strategy is dictated by its complexity. This helps explain well-known ICL phenomena, while offering novel predictions: e.g., we show a superlinear trend in the timescale for transitioning from generalization to memorization as task diversity increases. Overall, our work advances an explanatory and predictive account of ICL grounded in tradeoffs between strategy loss and complexity.