Search for: All records

Diffusion models have become the go-to method for text-to-image generation, producing high-quality images from noise through a process called reverse diffusion. Understanding the dynamics of the reverse diffusion process is crucial in steering the generation and achieving high sample quality. However, the inner workings of diffusion models is still largely a mystery due to their black-box nature and complex, multi-step generation process. Mechanistic Interpretability (MI) techniques, such as Sparse Autoencoders (SAEs), aim at uncovering the operating principles of models through granular analysis of their internal representations. These MI techniques have been successful in understanding and steering the behavior of large language models at scale. However, the great potential of SAEs has not yet been applied toward gaining insight into the intricate generative process of diffusion models. In this work, we leverage the SAE framework to probe the inner workings of a popular text-to-image diffusion model, and uncover a variety of human-interpretable concepts in its activations. Interestingly, we find that even before the first reverse diffusion step is completed, the final composition of the scene can be predicted surprisingly well by looking at the spatial distribution of activated concepts. Moreover, going beyond correlational analysis, we show that the discovered concepts have a causal effect on the model output and can be leveraged to steer the generative process. We design intervention techniques aimed at manipulating image composition and style, and demonstrate that (1) in early stages of diffusion image composition can be effectively controlled, (2) in the middle stages of diffusion image composition is finalized, however stylistic interventions are effective, and (3) in the final stages of diffusion only minor textural details are subject to change. 

est-time training (TTT) methods explicitly update the weights of a model to adapt to the specific test instance, and they have found success in a variety of settings, including most recently language modeling and reasoning. To demystify this success, we investigate a gradient-based TTT algorithm for in-context learning, where we train a transformer model on the in-context demonstrations provided in the test prompt. Specifically, we provide a comprehensive theoretical characterization of linear transformers when the update rule is a single gradient step. Our theory (i) delineates the role of alignment between pretraining distribution and target task, (ii) demystifies how TTT can alleviate distribution shift, and (iii) quantifies the sample complexity of TTT including how it can significantly reduce the eventual sample size required for in-context learning. As our empirical contribution, we study the benefits of TTT for TabPFN, a tabular foundation model. In line with our theory, we demonstrate that TTT significantly reduces the required sample size for tabular classification (3 to 5 times fewer) unlocking substantial inference efficiency with a negligible training cost. 

Large-scale channel prediction, i.e., estimation of the pathloss from geographical/morphological/building maps, is an essential component of wireless network planning. Ray tracing (RT)-based methods have been widely used for many years, but they require significant computational effort that may become prohibitive with the increased network densification and/or use of higher frequencies in B5G/6G systems. In this paper, we propose a data-driven, model-free pathloss map prediction (PMP) method, called PMNet. PMNet uses a supervised learning approach: it is trained on a limited amount of RT data and map data. Once trained, PMNet can predict pathloss over location with high accuracy (an RMSE level of 10−2 ) in a few milliseconds. We further extend PMNet by employing transfer learning (TL). TL allows PMNet to learn a new network scenario quickly ( ×5.6 faster training) and efficiently (using ×4.5 less data) by transferring knowledge from a pre-trained model, while retaining accuracy. Our results demonstrate that PMNet is a scalable and generalizable ML-based PMP method, showing its potential to be used in several network optimization applications. 

Diffusion models have established new state of the art in a multitude of computer vision tasks, in- cluding image restoration. Diffusion-based inverse problem solvers generate reconstructions of ex- ceptional visual quality from heavily corrupted measurements. However, in what is widely known as the perception-distortion trade-off, the price of perceptually appealing reconstructions is often paid in declined distortion metrics, such as PSNR. Distortion metrics measure faithfulness to the observation, a crucial requirement in inverse problems. In this work, we propose a novel framework for inverse problem solving, namely we assume that the observation comes from a stochastic degra- dation process that gradually degrades and noises the original clean image. We learn to reverse the degradation process in order to recover the clean image. Our technique maintains consistency with the original measurement throughout the reverse process, and allows for great flexibility in trading off perceptual quality for improved distortion metrics and sampling speedup via early-stopping. We demonstrate the efficiency of our method on different high-resolution datasets and inverse problems, achieving great improvements over other state-of-the-art diffusion-based methods with respect to both perceptual and distortion metrics. Source code and pre-trained models will be released soon. 

Indoor localization is a challenging task. Compared to outdoor environments where GPS is dominant, there is no robust and almost-universal approach. Recently, machine learning (ML) has emerged as the most promising approach for achieving accurate indoor localization. Nevertheless, its main challenge is requiring large datasets to train the neural networks. The data collection procedure is costly and laborious, requiring extensive measurements and labeling processes for different indoor environments. The situation can be improved by Data Augmentation (DA), a general framework to enlarge the datasets for ML, making ML systems more robust and increasing their generalization capabilities. This paper proposes two simple yet surprisingly effective DA algorithms for channel state information (CSI) based indoor localization motivated by physical considerations. We show that the number of measurements for a given accuracy requirement may be decreased by an order of magnitude. Specifically, we demonstrate the algorithms’ effectiveness by experiments conducted with a measured indoor WiFi measurement dataset: As little as 10% of the original dataset size is enough to get the same performance as the original dataset. We also showed that if we further augment the dataset with the proposed techniques, test accuracy is improved more than three-fold. 

Low-Earth orbit (LEO) satellite (SAT) networks exhibit ultra-wide coverage under time-varying SAT network topology. Such wide coverage makes the LEO SAT network support the massive IoT, however, such massive access put existing multiple access protocols ill-suited. To overcome this issue, in this paper, we propose a novel contention-based random access solution for massive IoT in LEO SAT networks. Not only showing the performance of our proposed approach (see, Table II), but we also discuss the issue of scalability of deep reinforcement learning (DRL) by showing the convergence behavior (see, Table III and IV). 

We explore the connection between outlier-robust high-dimensional statistics and non-convex optimization in the presence of sparsity constraints, with a focus on the fundamental tasks of robust sparse mean estimation and robust sparse PCA. We develop novel and simple optimization formulations for these problems such that any approximate stationary point of the associated optimization problem yields a near-optimal solution for the underlying robust estimation task. As a corollary, we obtain that any first-order method that efficiently converges to stationarity yields an efficient algorithm for these tasks. The obtained algorithms are simple, practical, and succeed under broader distributional assumptions compared to prior work. 

Deep neural networks have emerged as very successful tools for image restoration and reconstruction tasks. These networks are often trained end-to-end to directly reconstruct an image from a noisy or corrupted measurement of that image. To achieve state-of-the-art performance, training on large and diverse sets of images is considered critical. However, it is often difficult and/or expensive to collect large amounts of training images. Inspired by the success of Data Augmentation (DA) for classification problems, in this paper, we propose a pipeline for data augmentation for accelerated MRI reconstruction and study its effectiveness at reducing the required training data in a variety of settings. Our DA pipeline, MRAugment, is specifically designed to utilize the invariances present in medical imaging measurements as naive DA strategies that neglect the physics of the problem fail. Through extensive studies on multiple datasets we demonstrate that in the low-data regime DA prevents overfitting and can match or even surpass the state of the art while using significantly fewer training data, whereas in the high-data regime it has diminishing returns. Furthermore, our findings show that DA can improve the robustness of the model against various shifts in the test distribution.