Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks334 from public data
- The Split Bregman Method for L1-Regularized Problems4,507
This paper proposes a “split Bregman” method, which can solve a very broad class of L1-regularized problems, and applies this technique to the Rudin-Osher-Fatemi functional for image denoising and to a compressed sensing problem that arises in magnetic resonance imaging.
- A Watermark for Large Language Models1,108
A statistical test for detecting the watermark with interpretable p-values is proposed, and an information-theoretic framework for analyzing the sensitivity of the watermarks is derived.
- 926
- Fast Alternating Direction Optimization Methods837
This paper considers accelerated variants of two common alternating direction methods: the alternating direction method of multipliers (ADMM) and the alternating minimization algorithm (AMA), of the form first proposed by Nesterov for gradient descent methods.
- Baseline Defenses for Adversarial Attacks Against Aligned Language Models818
It is found that the weakness of existing discrete optimizers for text, combined with the relatively high costs of optimization, makes standard adaptive attacks more challenging for LLMs.
- Test-Time Prompt Tuning for Zero-Shot Generalization in Vision-Language Models612
Test-time prompt tuning (TPT) is proposed, a method that can learn adaptive prompts on the fly with a single test sample and performs on par with the state-of-the-art approaches that use additional training data.
- Diffusion Art or Digital Forgery? Investigating Data Replication in Diffusion Models560
This work studies image retrieval frameworks that enable us to compare generated images with training samples and detect when content has been replicated, and identifies cases where diffusion models, including the popular Stable Diffusion model, blatantly copy from their training data.
- Universal Guidance for Diffusion Models550
This work proposes a universal guidance algorithm that enables diffusion models to be controlled by arbitrary guidance modalities without the need to retrain any use-specific components.
- Geometric Applications of the Split Bregman Method: Segmentation and Surface Reconstruction492
The primary purpose of this paper is to examine the effectiveness of “Split Bregman” techniques for solving image segmentation problems, and to compare this scheme with more conventional methods.
- Dataset Security for Machine Learning: Data Poisoning, Backdoor Attacks, and Defenses438
The goal of this work is to systematically categorize and discuss a wide range of dataset vulnerabilities and exploits, approaches for defending against these threats, and an array of open problems in this space.
- BLIP3-o: A Family of Fully Open Unified Multimodal Models-Architecture, Training and Dataset394
A novel approach that employs a diffusion transformer to generate semantically rich CLIP image features, in contrast to conventional VAE-based representations is introduced, which yields both higher training efficiency and improved generative quality.
- Quantized Precoding for Massive MU-MIMO367
This paper investigates the problem of downlink precoding for a narrowband massive MU-MIMO system with low-resolution digital-to-analog converters at the base station and proposes novel nonlinear precoding algorithms that significantly outperform linear precoders at the cost of an increased computational complexity.
- 355
- Making an Invisibility Cloak: Real World Adversarial Attacks on Object Detectors334
A systematic study of adversarial attacks on state-of-the-art object detection frameworks, and a detailed study of physical world attacks using printed posters and wearable clothes, to quantify the performance of such attacks with different metrics.
- PhaseMax: Convex Phase Retrieval via Basis Pursuit295
This work formulate phase retrieval as a convex optimization problem, which it is shown that the dual problem to PhaseMax is basis pursuit, which implies that the phase retrieval can be performed using algorithms initially designed for sparse signal recovery.
- Tree-Ring Watermarks: Fingerprints for Diffusion Images that are Invisible and Robust279
This paper introduces a novel technique called Tree-Ring Watermarking that robustly fingerprints diffusion model outputs and demonstrates that this technique can be easily applied to arbitrary diffusion models, including text-conditioned Stable Diffusion, as a plug-in with negligible loss in FID.
- DCAN: Dual Channel-Wise Alignment Networks for Unsupervised Scene Adaptation276
Dual Channel-wise Alignment Networks (DCAN) are presented, a simple yet effective approach to reduce domain shift at both pixel-level and feature-level in deep neural networks for semantic segmentation.
- 268
- On the Reliability of Watermarks for Large Language Models251
This work finds that watermarks remain detectable even after human and machine paraphrasing, and considers a range of new detection schemes that are sensitive to short spans of watermarked text embedded inside a large document, and compares the robustness of watermarking to other kinds of detectors.
- Channel Charting: Locating Users Within the Radio Environment Using Channel State Information241
Channel charting (CC) is proposed, a novel framework in which a multi-antenna network element learns a chart of the radio geometry in its surrounding area that enables a range of new applications in UE localization, network planning, user scheduling, multipoint connectivity, hand-over, cell search, user grouping, and other cognitive tasks that rely on CSI and UE movement relative to the base station.
- 236
- Universal Adversarial Training215
This work proposes universal adversarial training, which models the problem of robust classifier generation as a two-player min-max game, and produces robust models with only 2× the cost of natural training.
- 184
- Towards Transferable Adversarial Attacks on Vision Transformers175
A dual attack framework is introduced, which contains a Pay No Attention (PNA) attack and a PatchOut attack, to improve the transferability of adversarial samples across different ViTs, and it is shown that skipping the gradients of attention during backpropagation can generate adversarial examples with high transferability.
- Strong Data Augmentation Sanitizes Poisoning and Backdoor Attacks Without an Accuracy Tradeoff160
It is found that strong data augmentations, such as mixup and CutMix, can significantly diminish the threat of poisoning and backdoor attacks without trading off performance.
- Deep k-NN Defense Against Clean-Label Data Poisoning Attacks155
This work proposes a simple, yet highly-effective Deep k-NN defense against both feature collision and convex polytope clean- label attacks on the CIFAR-10 dataset and shows that current clean-label poisoning attack strategies can be annulled, and serves as a strong yet simple-to-implement baseline defense to test future clean- Label poisoning attacks.
- NEFTune: Noisy Embeddings Improve Instruction Finetuning151
It is shown that language model finetuning can be improved, sometimes dramatically, with a simple augmentation, and even powerful models further refined with RLHF such as LLaMA-2-Chat benefit from additional training with NEFTune.
- Decentralized Baseband Processing for Massive MU-MIMO Systems145
This paper proposes a novel decentralized baseband processing architecture that alleviates bottlenecks by partitioning the BS antenna array into clusters, each associated with independent radio-frequency chains, analog and digital modulation circuitry, and computing hardware.
- Speedy-Splat: Fast 3D Gaussian Splatting with Sparse Pixels and Sparse Primitives140
This work identifies and addresses two key inefficiencies in 3D-GS to substantially improve rendering speed and introduces a novel pruning technique and integrates it into the training pipeline, significantly reducing model size and training time while further raising rendering speed.
- PUP 3D-GS: Principled Uncertainty Pruning for 3D Gaussian Splatting133
A principled sensitivity pruning score that preserves visual fidelity and foreground details at significantly higher compression ratios than existing approaches is proposed and a multi-round prune-refine pipeline that can be applied to any pretrained 3D-GS model without changing its training pipeline is proposed.
- FLAG: Adversarial Data Augmentation for Graph Neural Networks133
This work proposes a simple but effective solution, FLAG (Free Large-scale Adversarial Augmentation on Graphs), which iteratively augments node features with gradient-based adversarial perturbations during training, and boosts performance at test time.
- 132
- 130
- 128
- Cramming: Training a Language Model on a Single GPU in One Day123
This work investigates the downstream performance achievable with a transformer-based language model trained completely from scratch with masked language modeling for a single day on a single consumer GPU, and investigates why scaling down is hard, and which modifications actually improve performance in this scenario.
- Robust Optimization as Data Augmentation for Large-scale Graphs118
This work proposes FLAG (Free Large-scale Adversarial Augmentation on Graphs), which iteratively augments node features with gradient-based adversarial perturbations during training, which universally works in node classification, link prediction, and graph classification tasks.
- 118
- ACE: Adapting to Changing Environments for Semantic Segmentation110
This work presents ACE, a framework for semantic segmentation that dynamically adapts to changing environments over time, and introduces a memory that stores feature statistics from previously seen domains, thus preventing catastrophic forgetting.
- 1-bit Massive MU-MIMO Precoding in VLSI106
This paper designs VLSI architectures that enable efficient 1-bit precoding for massive MU-MIMO systems, in which hundreds of antennas serve tens of user equipments, and presents corresponding field-programmable gate array (FPGA) reference implementations to demonstrate that 1- bit precoding enables reliable and high-rate downlink data transmission in practical systems.
- What do Vision Transformers Learn? A Visual Exploration104
The obstacles to performing visualizations on ViTs are addressed, and it is shown that ViTs maintain spatial information in all layers except the final layer, and that the last layer most likely discards the spatial information and behaves as a learned global pooling operation.
- Can Neural Nets Learn the Same Model Twice? Investigating Reproducibility and Double Descent from the Decision Boundary Perspective89
It is seen that decision boundary reproducibility depends strongly on model width, and how these observations re-late to the theory of double descent phenomena in convex models is discussed.
- 89
- 88
- PhasePack: A Phase Retrieval Library87
The purpose of PhasePack is to create a common software interface for a wide range of phase retrieval algorithms and to provide a common testbed using both synthetic data and empirical imaging datasets.
- Automated Inference with Adaptive Batches87
This work proposes alternative “big batch” SGD schemes that adaptively grow the batch size over time to maintain a nearly constant signal-to-noise ratio in the gradient approximation.
- Transfer Learning with Deep Tabular Models84
This work demonstrates that upstream data gives tabular neural networks a decisive advantage over widely used GBDT models, and proposes a realistic medical diagnosis benchmark for tabular transfer learning and a how-to guide for using upstream data to boost performance with a variety oftabular neural network architectures.
- 84
- Nonlinear 1-bit precoding for massive MU-MIMO with higher-order modulation79
This paper develops novel algorithms for downlink precoding in massive MU-MIMO systems with 1-bit DACs that support higher-order modulation schemes such as 8-PSK or 16-QAM and presents low-complexity nonlinear precoding algorithms that achieve low error rates when combined with blind or training-based channel-estimation algorithms at the user equipment.
- 77
- Certified Neural Network Watermarks with Randomized Smoothing72
This paper shows that their watermark is guaranteed to be unremovable unless the model parameters are changed by more than a certain l2 threshold, and is also empirically more robust compared to previous watermarking methods.
- 71
- 65
- Siamese Neural Networks for Wireless Positioning and Channel Charting64
A unified architecture based on Siamese networks that can be used for supervised UE positioning and unsupervised channel charting and enables semisupervised positioning, where only a small set of location information is available during training.
- Exploring and Exploiting Decision Boundary Dynamics for Adversarial Robustness62
This work proposes Dynamics-aware Robust Training (DyART), which encourages the decision boundary to engage in movement that prioritizes increasing smaller margins, allowing for more targeted and effective robustness improvement.
- 62
- Joint Channel Estimation and Data Detection in Cell-Free Massive MU-MIMO Systems58
A joint channel estimation and data detection algorithm for densely-populated cell-free massive multiuser (MU) multiple-input multiple-output (MIMO) systems, which reduces the channel training overhead caused by the presence of hundreds of simultaneously transmitting user equipments (UEs).
- How Much Data Are Augmentations Worth? An Investigation into Scaling Laws, Invariance, and Implicit Regularization56
It is found that in out-of-distribution testing scenarios, augmentations which yield samples that are diverse, but inconsistent with the data distribution can be even more valuable than additional training data.
- 55
- Adaptive Relaxed ADMM: Convergence Theory and Practical Implementation51
The proposed adaptive relaxed ADMM (ARADMM) is derived by assuming a Barzilai-Borwein style linear gradient and automatically tunes the key algorithm parameters to achieve optimal performance without user oversight.
- 45
- 43
- Son of Zorn's lemma: Targeted style transfer using instance-aware semantic segmentation43
A method for targeted style transfer that simultaneously segments and stylizes single objects selected by the user, and uses a Markov random field model to smooth and anti-alias outlier pixels near object boundaries, so that stylized objects naturally blend into their surroundings.
- Improving Channel Charting with Representation -Constrained Autoencoders42
This paper demonstrates that autoencoder (AE)-based CC can be augmented with side information that is obtained during the CSI acquisition process and shows that such representation-constrained AEs recover the global geometry of the learned channel charts, which enables CC to perform approximate positioning without global navigation satellite systems or supervised learning methods that rely on extensive and expensive measurement campaigns.
- Efficient Distributed SGD with Variance Reduction42
This paper proposes a highly parallel method, CentralVR, that uses error corrections to reduce the variance of SGD gradient updates, and scales linearly with the number of worker nodes, and compares it to state-of-the-art parallel stochastic optimization methods on a variety of models and datasets.
- Adversarial attacks on machine learning systems for high-frequency trading41
This work introduces new attacks specific to this domain with size constraints that minimize attack costs and investigates the feasibility of realistic adversarial attacks in which an adversarial trader fools automated trading systems into making inaccurate predictions.
- 40
- Truth or backpropaganda? An empirical investigation of deep learning theory38
It is proved the widespread existence of suboptimal local minima in the loss landscape of neural networks, and it is shown that small-norm parameters are not optimal for generalization and rank does not correlate with generalization or robustness in a practical setting.
- 38
- 38
- Global minimization of Markov random fields with applications to optical flow36
This paper presents a variational functional lifting technique that can be viewed as a generalization of previous works by Pock et.
- Finite-Alphabet MMSE Equalization for All-Digital Massive MU-MIMO mmWave Communication35
FAME, short for finite-alphabet minimum mean-square error (MMSE) equalization, is introduced, which is able to significantly outperform a naïve quantization of the linear MMSE matrix.
- 34
- 34
- 33
- 32
- 32
- 32
- 31
- 31
- A New Rank Constraint on Multi-view Fundamental Matrices, and Its Application to Camera Location Recovery30
A novel rank constraint on collections of fundamental matrices in multi-view settings is introduced and it is shown that in general, with the selection of proper scale factors, a matrix formed by stacking fundamentalMatrices between pairs of images has rank 6.
- 29
- 29
- Bring Your Own Data! Self-Supervised Evaluation for Large Language Models27
This work proposes a framework for self-supervised evaluation of LLMs by analyzing their sensitivity or invariance to transformations on the input text, and demonstrates self- supervised evaluation strategies for measuring closed-book knowledge, toxicity, and long-range context dependence.
- 26
- 26
- Biconvex Relaxation for Semidefinite Programming in Computer Vision26
This work proposes a general framework to approximately solve large-scale semidefinite problems (SDPs) at low complexity, and transforms a general SDP into a specific biconvex optimization problem, which can then be solved in the original, low-dimensional variable space atLow complexity.
- 26
- 25
- 25
- 24
- A Deep Dive into Dataset Imbalance and Bias in Face Identification23
This work thoroughly explores the effects of each kind of imbalance possible in face identification, and discusses other factors which may impact bias in this setting.
- 23
- 22
- 22
- 21
- Are Adversarial Examples Created Equal? A Learnable Weighted Minimax Risk for Robustness under Non-uniform Attacks21
A weighted minimax risk optimization that defends against non-uniform attacks, achieving robustness against adversarial examples under perturbed test data distributions and significantly improves state-of-the-art adversarial accuracy under non- uniform attacks.
- 20
- 20
- Object Recognition as Next Token Prediction19
An approach to pose object recognition as next token prediction as next token prediction is presented, and a simple strategy to construct a compact decoder is proposed by simply discarding the intermediate blocks of a pretrained language model.
- 18
- 18
- Witchcraft: Efficient PGD Attacks with Random Step Size17
This paper proposes a variant of Projected Gradient Descent (PGD) that uses a random step size to improve performance without resorting to expensive random restarts, which makes crafting attacks more economical, which is important in situations like adversarial training where attacks need to be crafted in real time.
- 17
- 16
- 15
- 15
- 15
- 14
- 14
- 14
- 13
- 13
- 11
- 11
- 11
- High-Bandwidth Spatial Equalization for mmWave Massive MU-MIMO With Processing-in-Memory10
This brief explores two different finite-alphabet equalization hardware implementations that tightly integrate the memory and processing elements: a parallel array of multiply-accumulate (MAC) units and a bit-serial processing-in-memory (PIM) architecture.
- 10
- 9
- Hierarchical Point Attention for Indoor 3D Object Detection9
This work proposes two novel attention operations as generic hierarchical designs for point-based transformer detectors that builds multi-scale tokens from a single-scale input feature to enable more fine-grained feature learning.
- 9
- 9
- 9
- 9
- 9
- 9
- FPGA design of approximate semidefinite relaxation for data detection in large MIMO wireless systems9
- 8
- 8
- 8
- Prepare for the Worst: Generalizing across Domain Shifts with Adversarial Batch Normalization8
This work adapts adversarial training by adversarially perturbing feature statistics, rather than image pixels, to produce models that are robust to domain shift, and significantly improves the performance of ResNet-50 on ImageNet-C, Stylized-ImageNet, and Image net-Instagram over standard training practices.
- 8
- 8
- 8
- 8
- 7
- 7
- Preparing for the Worst: Making Networks Less Brittle with Adversarial Batch Normalization7
This work adapts adversarial training by adversarially perturbing feature statistics, rather than image pixels, to produce models that are robust to domain shift, and significantly improves the performance of ResNet-50 on ImageNet-C, Stylized-ImageNet, and Image net-Instagram over standard training practices.
- 7
- 7
- Strong Baseline Defenses Against Clean-Label Poisoning Attacks7
This work proposes a set of simple, yet highly-effective defenses against targeted clean-label poisoning attacks, and shows that current clean- label poisoning attack strategies can be annulled, and serve as strong but simple-to-implement baseline defense for which to test future clean- Label poisoning attacks.
- 7
- 7
- 6
- 6
- 6
- 6
- 6
- 5
- 5
- 5
- 5
- 5
- 5
- 5
- 5
- 5
- 5
- 5
- 5
- 4
- 4
- 4
- 4
- 4
- 4
- Adaptive Learning Rates with Maximum Variation Averaging4
An adaptive learning rate rule is proposed in which the running mean squared gradient is replaced by a weighted mean, with weights chosen to maximize the estimated variance of each coordinate, which gives a worst-case estimate for the local gradient variance.
- 4
- 4
- 4
- 4
- 4
- 4
- 4
- 3
- 3
- 3
- 3
- 3
- 3
- 3
- 3
- 3
- MaxVA: Fast Adaptation of Step Sizes by Maximizing Observed Variance of Gradients3
An adaptive learning rate principle is proposed, in which the running mean of squared gradient in Adam is replaced by a weighted mean, with weights chosen to maximize the estimated variance of each coordinate, which results in a faster adaptation to the local gradient variance, which leads to more desirable empirical convergence behaviors than Adam.
- 3
- 3
- 3
- 3
- 3
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- TransFIRA: Transfer Learning for Face Image Recognizability Assessment1
TransFIRA is introduced, a unified, geometry-driven framework for recognizability assessment that is encoder-specific, accurate, interpretable, and extensible across modalities, significantly advancing FIQA in accuracy, explainability, and scope.
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Universal Pyramid Adversarial Training for Improved ViT Performance–
This work proposes Universal Pyramid Adversarial training, where a single pyramid adversarial pattern is shared across the whole dataset instead of the sample-wise patterns, and is the first to find that universal adversarial training can be leveraged to improve clean model performance.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-16. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.