Gradient_Descent
-
DoctorAgents: an agentic framework to iteratively refine AutoML pipeline for small clinical temporal data
2026-08-05 19:52:55Ruilin Wang, Bo-Hong Wang, Elizabeth Kourbatski, Jun Bai, Hegang Chen, Ziyang Song, Gilles Boire, Marie Hudson, Yue LiAbstract
Clinical machine learning (ML) has the potential to support high-stakes medical decision-making, but reliable deployment is often constrained by scarce, heterogeneous, and temporal complexity. Developing effective ML pipelines for such data remains time-consuming and error-prone, while existing automated machine learning (AutoML) systems only partially address this challenge because they largely rely on brute-force search over predefined spaces and lack explicit reasoning and memory. We therefore reformulate AutoML for small clinical data from exhaustive search to reasoning-driven refinement. We propose DoctorAgents, an agentic AI framework that autonomously constructs and optimizes end-to-end ML pipelines through specialized large language model (LLM) agents for generation, validation, and refinement. DoctorAgents backpropagates natural-language feedback through textual gradient descent to perform targeted updates without exhaustive search. Experiments across diverse clinical tasks show that DoctorAgents consistently outperforms established AutoML baselines while producing more interpretable task-specific representations.
Abstract (translated)
URL
https://arxiv.org/abs/2608.05375
PDF
-
SVI-DAG: A Structured Variational Inference Approach to Bayesian Causal Discovery
2026-08-05 14:56:14Shrenik ZinageAbstract
Bayesian causal discovery seeks to determine the posterior distribution of causal theories, which are interpreted as directed acyclic graphs (DAGs) that explain the observed data. The resulting posterior allows systematic reasoning regarding epistemic uncertainty within these theories. Nonetheless, finding such graphs is difficult due to identifiability problems and limited observational data. Furthermore, precisely approximating posterior over graphs is challenging given vast range of potential DAGs. Recent Bayesian approaches have addressed some of these challenges, yet they remain limited as they fail to encode dependencies between edges, and lack principled ways to incorporate domain knowledge as inductive biases during the search process. To overcome these limitations, we propose SVI-DAG, a structured variational inference approach to Bayesian causal discovery using observational data and prior beliefs that uses normalizing flows to model dependencies between edges, supporting expressive and multimodal posterior learning over DAGs. To mitigate mode seeking behaviour in evidence lower bound optimization and promote mode coverage, we use stein variational gradient descent to update the node potentials using a kernel in acyclicity space. We evaluate SVI-DAG against 5 state-of-the-art Bayesian DAG learning methods and demonstrate superior performance in uncertainty quantification while remaining competitive in terms of structural accuracy.
Abstract (translated)
URL
https://arxiv.org/abs/2608.04930
PDF
-
MESH: Memory-Efficient Sinkhorn Optimization for Mixture-of-Experts Training
2026-08-05 03:23:31Masato FujitakeAbstract
Memory-efficient matrix optimizers such as Sinkhorn gradient descent remove most AdamW optimizer state for dense Transformer matrices, but direct application to Mixture-of-Experts (MoE) training is unreliable. We study this failure in a controlled 110M-parameter nanowhale DeepSeek-style MoE pretraining setting. A SAGE/Sinkhorn hybrid reduces optimizer state from 0.883GB to 0.331GB but degrades evaluation loss to 3.8265, far above the AdamW baselines observed in the same setup (3.58--3.64 across the seeds we study). We show that routed MoE expert matrices are the dominant failure point: their gradients are conditional, temporally varying, and poorly served by stateless Sinkhorn normalization. We propose MESH, a hidden-momentum Sinkhorn update for MoE experts. MESH restores a temporal first-moment signal through the gradient-buffer lifecycle, without storing the expert first moment as optimizer state. MESH is an optional block-preconditioned variant that adds a coarse neuron/block inverse-RMS multiplier. Across ablations, temporal smoothing before matrix normalization is the primary causal ingredient; block/neuron preconditioning can improve the memory-quality frontier, but is not established as universally necessary. In two additional seeds, MESH and MESH-B reduce optimizer-state memory by 62.5\% and peak PyTorch CUDA allocation by about 12.6\% relative to AdamW, with a modest evaluation-loss gap. Full-state diagnostic variants recover AdamW-like performance in ablations, supporting the conclusion that MoE experts need temporal smoothing, but not necessarily full coordinate-wise AdamW state.
Abstract (translated)
URL
https://arxiv.org/abs/2608.04407
PDF
-
SRAP: SVD-Refined Adversarial Perturbations for Imperceptible Face-Swap Defense
2026-08-04 09:48:46Sungwon Cho, Kwanghyun Ko, Myungjoo KangAbstract
Deepfake technologies pose increasing threats to facial privacy and identity security, motivating proactive defenses that protect facial images before misuse. Although adversarial perturbations generated by projected gradient descent (PGD) can disrupt the identity representations used by face-swapping models, their visual quality is degraded by two characteristics: perturbations are distributed broadly over the image, including identity-insensitive regions, and they contain visually salient high-frequency components. We analyze these spatial and spectral inefficiencies through identity-sensitivity estimation and the singular-value decomposition (SVD) of PGD perturbations. Our analysis shows that later singular components contain a disproportionate amount of high-frequency energy, while the leading components preserve most of the perturbation energy and defense utility. Based on these observations, we propose SRAP, which combines per-channel truncated SVD refinement with an identity-importance mask at every optimization step. The SVD refinement suppresses high-rank, high-frequency residuals, while the mask restricts perturbations to locations that strongly influence identity representations. Experiments on CelebA-HQ and VGGFace2-HQ demonstrate that SRAP substantially improves protected-image fidelity across all reported metrics while maintaining competitive identity-disruption performance, yielding a favorable trade-off between face-swap defense and visual imperceptibility.
Abstract (translated)
URL
https://arxiv.org/abs/2608.03395
PDF
-
NanoMorph-3D: An End-to-End Physics-Driven Unrolling Framework for Nanomaterial Reconstruction
2026-08-04 07:30:48Beiyuan Zhang, Hesong Li, Ziqi Wu, Ruiwen Shao, Ying FuAbstract
Precise 3D characterization of nanomaterials is essential for unlocking structure-property relationships. However, standard electron tomography is fundamentally limited by the missing wedge problem. Consequently, conventional algorithms suffer from severe geometric distortions, a challenge further complicated by pervasive noise interference. Current learning-based methods either rely on physics-blind post-processing or employ end-to-end architectures constrained by local receptive fields, failing to capture complex 3D topologies. We propose NanoMorph-3D, a unified end-to-end framework grounded in a comprehensive Nanomorphological Taxonomy. Powered by a large-scale synthetic dataset explicitly modeling non-linear electron attenuation, we design a Physics-Driven Unrolled Network mapping proximal gradient descent into a learnable architecture. To capture complex internal topologies, we formulate a hierarchical attention mechanism with Physics-Normalization for long-range 3D dependencies and scale invariance. Crucially, our Dual-Domain strategy leverages Sinusoidal Attention to explicitly model physical projection trajectories, enforcing strict sinogram consistency to mitigate missing wedge artifacts. Finally, an unsupervised dual-stream mechanism bridges the simulation-to-reality gap. Experiments demonstrate NanoMorph-3D reconstructs diverse topologies with superior fidelity and speed.
Abstract (translated)
URL
https://arxiv.org/abs/2608.03257
PDF
-
From Digital to Physical Reservoir Computing: Co-Optimizing Soft Robotic Reservoirs via Dynamics Matching
2026-08-01 07:21:23Nicola Visentin, Maximilian St\"olzle, Mariano Ram\'irez Montero, Francesco Braghin, Daniela Rus, Cosimo Della SantinaAbstract
Soft robotic substrates are promising for Physical Reservoir Computing (PRC) because their compliant nonlinear dynamics can provide temporal memory, high-dimensional state transformations, and efficient inference. However, physical reservoirs are often adopted as-is rather than pretrained or co-optimized, potentially limiting soft robotic PRC performance relative to digital reservoirs. We investigate whether a physical reservoir can instead be pretrained against high-performing digital reference dynamics. Our formulation jointly optimizes physical parameters, a diffeomorphic physical-reference state map, and feedforward-feedback control using a differentiable physical model and an acceleration-level equation-error objective that avoids temporal integration. As a proof of concept, we instantiate the formulation with simulated soft robots, a Random Oscillators Network (RON) reference, and parallel multi-start gradient descent. We evaluate the optimized reservoirs on classification (sMNIST and ADIAC) and forecasting (Mackey-Glass and Lorenz96) tasks across four reservoir dimensions. Compared with unoptimized soft robot reservoirs, the optimized reservoirs achieve a mean relative improvement of 33.7% across all tasks and datasets, while remaining close to the digital reference. These results demonstrate the feasibility of dynamics-level co-optimization for the simulated soft robotic reservoirs considered here.
Abstract (translated)
URL
https://arxiv.org/abs/2608.00484
PDF
-
Shared Symbolic Backbones for Physically Consistent Multi-Output Symbolic Regression
2026-07-29 06:47:36Manuel RodriguezAbstract
Symbolic regression provides analytical expressions, but it is usually applied one output at a time. This is limiting in process systems, where state variables are often coupled through shared physical parameters. Independent symbolic regression can give accurate individual equations that are difficult to interpret as one model. We present a neuro-evolutionary symbolic regression method for coupled multi-output systems. The method searches for a shared symbolic backbone: a set of latent symbolic units that is discovered once and reused by several outputs through sparse additive or multiplicative read-outs. The discrete model structure is evolved by mutation and crossover, whereas the continuous parameters are tuned by gradient descent and inherited by the offspring. The method is assessed on a set of benchmarks with known ground truth and on a hydrothermal liquefaction yield case. The results show that coupling is not a general route to lower prediction error. Its main contribution is the enforcement and diagnosis of cross-output consistency when a physically shared factor is embedded in a latent expression and is weakly identifiable from the data. This occurs for Langmuir-Hinshelwood and site-coverage denominators, for which independent PySR does not close the consistency gap or recover the same shared form. Conversely, when each output is already identifiable, as in the Van de Vusse benchmark, independent symbolic regression matches or improves the coupled model. The proposed framework, rather than a general purpose predictor, is a structured shared-mechanism extractor. Its value is highest when the target structure is sparse, shared, weakly identifiable or constrained by closure.
Abstract (translated)
URL
https://arxiv.org/abs/2607.26528
PDF
-
Quotient Dynamics, Effective Curvature, and Implicit Bias in Positive Quadratic Networks
Abstract
Positive quadratic networks admit the low-rank representation f_U(x)=x^top UU^top x, where Uinmathbb{R}^{dtimes r} is identifiable only up to right orthogonal multiplication, representing a rank-r PSD matrix Q=UU^top. We study how this quotient structure governs training dynamics, curvature, recovery, and interpolation bias. On the full-column-rank stratum, we identify mathbb{R}^{dtimes r}_*/O(r) with the rank-r PSD manifold. For smooth objectives L(U)=ell(UU^top), the Euclidean factor gradient is horizontal. Thus, factor gradient flow projects exactly to quotient Riemannian gradient flow, while finite-step gradient descent induces an exact congruence recursion for the predictor. For quadratic regression, we derive the effective Hessian at interpolators as the empirical measurement Gram form restricted to the tangent space relative to the quotient metric. Under Gaussian rank-one measurements, we compute population curvature, prove uniform deviation bounds for the empirical normal operator, construct a spectral initializer, and establish local exponential convergence for gradient flow and linear convergence for small-step descent. Recovery guarantees are explicit but conservative due to reliance on full-space second-moment control. In underdetermined commuting regimes, factor gradient flow becomes an exact entropy mirror flow in joint spectral coordinates. Strictly positive initializations converge to Bregman projections onto the interpolation set. With isotropic initialization q(0)=varepsilon^2mathbf{1}, predictors approach the minimum-trace solution set as varepsilondownarrow0, resolving nonuniqueness via weighted entropy within the invariant joint spectral algebra. Finite-step descent selects interpolants differing from continuous-time Bregman projections by O(eta). Numerical experiments verify these quotient identities, curvature predictions, recovery behaviors, and selection laws.
Abstract (translated)
URL
https://arxiv.org/abs/2607.25624
PDF
-
Extracting Voice Styles from Frozen TTS Models via Gradient-Based Inverse Optimization
Abstract
Some text-to-speech systems ship a synthesis model and preset style vectors but not the reference encoder that turns audio into such a vector. The model still accepts a style vector; a user with a voice of their own cannot produce one. We solve for that input directly, inverting the released pipeline by gradient descent: every weight stays frozen and only the style vector is optimized, against time-pooled WavLM statistics of one recording. Because the objective discards the time axis, the synthesized text may differ from the recording, so no transcript and no alignment are needed. On 154 speakers from two corpora, ECAPA-TDNN similarity rises from 0.132 to 0.413 and ResNet from 0.099 to 0.401, improving for every speaker; a verifier at its equal-error point accepts 53% of the recovered voices as the target, against 1% for the presets they start from.
Abstract (translated)
URL
https://arxiv.org/abs/2607.25351
PDF
-
Scale Weight Decay and Train Better
2026-07-26 17:55:40Anuj ApteAbstract
The discovery of scaling laws has motivated training neural networks on ever increasing quantities of data. This is typically done with a constant decoupled weight decay which causes the network weights to shrink steadily over the course of training. Taking inspiration from the Robbins--Monro conditions, we propose to scale weight decay by the fraction of the peak learning rate $\eta/\eta_{\max}$. We prove that this scaled weight decay preserves the asymptotic stationarity guarantees of the corresponding unregularized methods for both stochastic gradient descent and the non-Euclidean spectral optimizer Muon, thereby avoiding the additional asymptotic bias introduced by constant decoupled weight decay. This retains the stability benefits of weight decay without changing the asymptotic optimization target. Using a steady-state analysis, we explain why under standard weight decay the weight norm shrinks steadily as training proceeds, whereas under scaled weight decay it settles to a roughly constant value. When applied to the training of mixture-of-experts models, Muon with scaled weight decay (Muon-SW) consistently outpaces Muon with identical hyperparameters, reaching the same validation loss $\mathbf{30\%}$ faster at our largest scale across models from $72 - 930$ million parameters trained at $\sim 600$ tokens per active parameter. If this trend continues to hold, the method promises to substantially accelerate the pre-training of frontier models while requiring only a few lines of code to implement.
Abstract (translated)
URL
https://arxiv.org/abs/2607.23777
PDF
-
Exact Network Surgery: Functional Invariance and Gradient Plasticity in Reactive Computational Graphs
2026-07-18 00:48:24Abdallah Khemais (ISITCOM, University of Sousse)Abstract
Function-preserving network growth techniques such as Net2Net and progressive stacking expand a model's capacity without destroying its learned function, but existing formulations either tolerate numerical perturbations or require a full rebuild of the training program. We formalize Exact Network Surgery: the in-place insertion of a residual block into a live computational graph such that (i) the network function is preserved -- bit-exactly under explicit floating-point hypotheses -- and (ii) inserted parameters remain trainable immediately after insertion. We prove an identity-morphism theorem for gated residual blocks, a structural-locality theorem showing that a reactive invalidation engine recomputes exactly the downstream cone of the insertion point, leaving every other node's value and optimizer state untouched, and an escape-from-initialization proposition showing that the Gradient Shadowing gate alpha, initialized at zero over a randomly initialized branch, receives a generically non-zero gradient at insertion time. We identify a degenerate configuration -- zero-initialized output projections combined with a zero gate -- that is an exact saddle point gradient descent cannot escape. Every claim is validated on the reference implementation in NeuroDSL, a reactive graph engine in Julia: grafting is bit-exact on every logit tested (0 mismatches out of 1600); the gate escapes zero at the first optimizer step and unlocks branch gradients at the second, exactly as predicted; the degenerate configuration exhibits gradients identically zero for the entire 600-step run; surgery cost tracks downstream cone size with r = 0.9992 while graft-plus-invalidation bookkeeping is constant (about 0.75 ms) across insertion depths; and training resumes bit-identically across a real process restart. A flagged preliminary appendix reports first single-seed observations on post-insertion gate dynamics.
Abstract (translated)
URL
https://arxiv.org/abs/2607.16568
PDF
-
In-context learning of closed form solution to simple linear regression task using transformer with linear self-attention
2026-07-17 10:28:27Katsuyuki HagiwaraAbstract
In-context learning is a remarkable property of transformers and has recently received a lot of interest. In many studies of in-context learning, it has been shown that transformers are capable of implementing solver for linear and non-linear regression problems, in which the most of them implement gradient descent algorithm. However, it is still unclear whether those implementations have actually been acquired through training. In this paper, we construct a transformer with linear self-attention, which in-context learns the least squares estimate in a simple regression task. The point here is that the closed form (analytical) solution is approximately obtained by using layer normalization rather than an approximate solution based on gradient descent algorithm. Then, we show an experimental example, in which our implementation is mainly used in the transformer trained with l1 regularization when the target output is the least squares estimate.
Abstract (translated)
URL
https://arxiv.org/abs/2607.15819
PDF
-
RoboTTT: Context Scaling for Robot Policies
2026-07-16 17:59:06Yunfan Jiang, Yevgen Chebotar, Ruijie Zheng, Fengyuan Hu, Yunhao Ge, Jimmy Wu, Tianyuan Dai, Scott Reed, Li Fei-Fei, Yuke Zhu, Linxi "Jim" FanAbstract
Recent robot foundation models operate with single-step or short-history visuomotor context. We introduce Test-Time-Training Robot Policies (RoboTTT), a robot model and training recipe that scale visuomotor context to 8K timesteps, three orders of magnitude beyond state-of-the-art policies, without growing inference latency. At this context length, we unlock new robot capabilities: one-shot in-context imitation from human video demonstrations, on-the-fly policy improvement, robustness to perturbations, and stronger performance on multi-stage, long-horizon tasks. We also observe, for the first time, steady gains in closed-loop performance as pretraining context length scales. At its core, RoboTTT integrates Test-Time Training into robot foundation models such as Vision-Language-Action policies, yielding a sequence model whose recurrent state consists of fast weights, parameters updated by gradient descent during both training and inference, compressing histories into weight space and retrieving contextual information for long-context conditioning. To scale training context length, the recipe combines sequence action forcing with truncated backpropagation through time. On challenging real-robot manipulation tasks, RoboTTT improves overall performance by 87% over the single-step context baseline and fully completes a five-minute, ten-stage assembly task, which no baseline ever does. RoboTTT trained with 8K-timestep context outperforms the same model pretrained with 1K timesteps by 62%, suggesting context length as a new scaling axis for robot foundation models. Videos are available at this https URL
Abstract (translated)
URL
https://arxiv.org/abs/2607.15275
PDF
-
Efficient Tuning Before Low-Bit Post-Training Quantization for Stochastic Gradient Descent-optimized Models
2026-07-13 10:24:36Peng Xia, Junbiao Pang, Muhammad Ayub SabirAbstract
Post-training quantization (PTQ) compresses deep neural networks for deployment under limited memory and computational budgets. However, low-bit (i.e., 2-bit or 4-bit) PTQ often suffers from substantial performance degradation. Most existing PTQ methods operate on an unconstrained full-precision (FP) model and primarily address quantization errors through post-hoc reconstruction. We argue that low-bit PTQ accuracy is limited not only by post-quantization error minimization, but also by the quantization-error tolerance of a FP model itself. In this paper, we propose Efficient Tuning Before Quantization (ETBQ), a pre-conditioning tuning stage for Stochastic Gradient Descent (SGD)-optimized models before PTQ. During tuning, the FP model is optimized under perturbations sampled from the error distributions of weight and activation quantization, guiding the model toward a loss-landscape region that is less sensitive to the subsequent PTQ. Unlike QAT, ETBQ does not train a fake-quantized deployment model, which is computationally and memory intensive. Instead, ETBQ outputs a FP model that can be used by any PTQ backend. Experiments on CIFAR-100, Tiny-ImageNet, ImageNet, and Cityscapes provide consistent evidence that ETBQ improves low-bit PTQ across diverse tasks. Under W2A4 settings, e.g., ETBQ improves over naive PTQ by 2.14\% top-1 accuracy on Tiny-ImageNet and by 5.80\% mIoU on Cityscapes. Code is available at this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2607.11359
PDF
-
Quantum Compressed Sensing CT Reconstruction Algorithm Based on Penalized Weighted Least Squares and Guided Total Variation
2026-07-12 04:48:08Yuwen Zhang, Yujie Liu, Ao Wang, Yikuang Yuluo, Shuangyang Zhong, Haijun Yu, Yixing HuangAbstract
Objective. Existing quadratic unconstrained binary optimization (QUBO)-based sparse-view computed tomography (CT) reconstruction neglects photon-counting statistics and anatomical heterogeneity. We address both limitations within the QUBO this http URL. We propose a quantum compressed-sensing CT method combining penalized weighted least squares (PWLS) and guided total variation (GTV). PWLS weights projection residuals by photon-count reliability, whereas GTV uses gradients from a prior image reconstructed by the simultaneous algebraic reconstruction technique (SART) to preserve edges and suppress noise in homogeneous regions. After binary encoding, both terms form a unified QUBO model. Experiments used four 40 times 40 CT images under a 10-view fan-beam geometry with Poisson noise. Comparisons included conventional reconstruction methods, QUBO variants, gradient descent, simulated annealing, and a D-Wave hybrid quantum-classical this http URL results. PWLS-GTV achieved the best reconstruction quality across all cases. In the representative chest case, it reached a peak signal-to-noise ratio (PSNR) of 36.64 dB, compared with 22.48 dB for SART, the best conventional baseline. GTV consistently outperformed conventional total variation. Simulated annealing and the D-Wave hybrid solver produced similar reconstructions, whereas gradient descent was ineffective. Repeated hybrid-solver runs showed stable this http URL. The framework incorporates photon-statistical weighting and structure-guided regularization into QUBO-based CT reconstruction without changing its quadratic form, providing a proof of concept for quantum-assisted sparse-view CT reconstruction.
Abstract (translated)
URL
https://arxiv.org/abs/2607.10566
PDF
-
Hybrid Least Squares/Gradient Descent Methods for MIONets
2026-07-08 03:59:25Jun Choi, Chang-Ock Lee, Minam MoonAbstract
In this paper, we propose an efficient hybrid least squares/gradient descent (LSGD) method for MIONets to accelerate training. This method generalizes the LSGD method for DeepONets. Since MIONet is the sum of the entrywise product of multiple branch networks and a trunk network, it can be viewed as a multilinear function with respect to the last layer parameters of each branch network. These sets of parameters can be optimized using the alternating least squares method, where we solve the LS system for a single branch network in turn. To handle the large-sized system matrix, we introduce Kronecker and Khatri-Rao products and tensor permutation matrices to factor the large matrix into small ones. Our method is compatible with a general type of $L^2$ loss with regularization terms for the last layer parameters of each branch, where linear operators can be applied to the MIONet output in each loss term.
Abstract (translated)
URL
https://arxiv.org/abs/2607.06976
PDF
-
Differentially Private Natural Gradient Descent
2026-07-07 05:49:08Pan Li, Kai Chen, Shuai Chang, Shengzhi Zhang, Peizhuo Lv, Jinwen HeAbstract
Under a fixed privacy budget, the utility of differentially private (DP) training is ultimately determined by its optimization efficiency. Standard first-order DP optimizers such as DP-SGD rely solely on local gradients and ignore the underlying loss curvature. This geometric blindness causes severe zigzagging in ill-conditioned landscapes, squandering precious privacy budgets on inefficient iterations. Practitioners are thus trapped in a bind: either stop training prematurely or inject massive per-step noise, both of which critically compromise final model utility. Natural Gradient Descent (NGD) resolves this by preconditioning gradients with curvature, aligning updates with the loss geometry and extracting more efficient signal from every noisy step, offering a principled pathway to break the privacy-utility bottleneck. Despite its theoretical appeal, directly integrating NGD with DP introduces fundamental challenges: curvature estimation itself consumes prohibitive privacy budgets, isotropic DP operations conflict with the anisotropic scaling of NGD, and the inverse curvature catastrophically amplify parameter updates in flat directions, causing training instability. We propose DP-NGD, a practical framework that systematically addresses these obstacles by decoupling curvature estimation from private data, reconciling isotropic DP constraints with anisotropic second-order optimization via a whitened-space mechanism, and dynamically clamping the curvature to stabilize training. Extensive experiments on standard benchmarks demonstrate that DP-NGD achieves state-of-the-art accuracy, breaking through the utility ceilings of first-order baselines while delivering up to a $10\times$ convergence speedup under the same privacy budget.
Abstract (translated)
URL
https://arxiv.org/abs/2607.05866
PDF
-
Learning Probabilistic Embeddings for Unsupervised Action Segmentation
2026-07-06 16:10:25Shuai Li, Duc Manh Vu, Juergen GallAbstract
This paper concerns the problem of unsupervised temporal action segmentation for long, untrimmed videos. Recent successful approaches follow a joint representation learning and clustering paradigm, where optimal transport (OT) is adopted to produce pseudo labels for learning frame representations. These approaches alternate between estimating pseudo labels using OT and optimizing the parameters with gradient descent during training, where OT is used for obtaining the final temporal action segmentation. A major limitation of these works is that they learn a deterministic embedding for frame representations. The iterative procedure between learning deterministic embeddings based on pseudo labels and estimating pseudo labels from the learned embedding can thus get quickly stuck in a local optimum. As an alternative, we thus propose to learn a probabilistic embedding for frame representations. The embeddings are modeled by Gaussian distributions and we sample from the distributions before estimating the pseudo labels. We evaluate our approach on several challenging temporal action segmentation datasets and achieve results comparable to, and in some cases, better than the state of the art. Compared to baselines with deterministic embeddings, our approach improves MoF up to 20.7\% and F1-score up to 19.0\%. Our code is available at this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2607.05263
PDF
-
The Map Behind the Flow: Finite-Step Gradient Descent as a Dynamical System
2026-07-06 12:28:24Thomas HofmannAbstract
Many phenomena of deep learning are dynamical: they concern not only which minima exist, but how gradient descent reaches, avoids, or selects among them. Edge-of-stability behavior, sharpness oscillations, catapult phases, balancing, and movement toward flatter representations are effects of the training map itself, and are poorly captured by the small-step gradient-flow limit. This paper studies fixed-step gradient descent as a discrete dynamical system in a hierarchy of exactly solvable models retaining basic structures of deep learning: depth, factorization, width, data coupling, activation, and stochasticity. The starting point is the balanced scalar reduction of a deep linear chain, giving a quartic loss and a cubic gradient map whose post-edge behavior is explicit. Under the natural large-depth scaling, this dynamics converges to a universal Ricker-type map. The edge of stability is therefore not a breakdown of optimization, but the first bifurcation of the training map. Embedding the scalar dynamics back into factored models turns these regimes into learning phenomena. Finite steps break conservation laws of gradient flow and contract factorization imbalance; residual oscillations move parameters toward flatter, more balanced representations. Wider linear networks produce a ladder of spectral edges, so the optimal learning rate can lie beyond the first edge. Data coupling, nonlinear activations, and stochastic targets preserve the same organizing principle: finite-step oscillations drive alignment, balancing, and representation selection. Thus the learning rate is not merely a numerical stability parameter. It is a structural parameter of the training dynamics, determining its attractors and shaping the representations gradient descent selects.
Abstract (translated)
URL
https://arxiv.org/abs/2607.04993
PDF
-
Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
2026-06-30 06:13:38Jiachun Li, David Simchi-LeviAbstract
Adaptive experiments for average treatment effects (ATE) require randomized allocations balancing valid inference with statistical efficiency. The oracle design is a covariate-dependent Neyman rule governed by unknown arm-conditional outcome variances. We investigate whether this sequential variance-estimation and allocation process can be amortized via in-context learning. We introduce Bayesian in-context experimenters: transformer policies trained to imitate a Bayesian posterior Neyman teacher. The teacher updates nonparametric beliefs over potential outcomes using experimental history to assign posterior Neyman treatment probabilities. This design converges to the oracle rule, supporting efficient ATE inference. Transformers constructively implement this mapping through attention-based sufficient statistics and projected gradient descent, imitating Bayesian updating for Gaussian-series priors. To address unknown outcome smoothness, we combine smoothness-indexed experimenters using a mixture-of-experts transformer. The gate acts as a hierarchical posterior over smoothness classes, concentrating on near-oracle experts. By bounding the complexity of the transformer class, we prove this amortized policy can be learned via empirical risk minimization using supervised pretraining. Experiments confirm accurate teacher imitation, adaptive allocation, and improved ATE precision over baselines.
Abstract (translated)
URL
https://arxiv.org/abs/2606.31184
PDF
