Search

Recent studies have demonstrated the strong empirical performance of diffusion models on discrete sequences across domains from natural language to biological sequence generation. For example, in the protein inverse folding task, conditional diffusion models have achieved impressive results in generating natural-like sequences that fold back into the original structure. However, practical design tasks often require not only modeling a conditional distribution but also optimizing specific task objectives. For instance, we may prefer protein sequences with high stability. To address this, we consider the scenario where we have pre-trained discrete diffusion models that can generate natural-like sequences, as well as reward models that map sequences to task objectives. We then formulate the reward maximization problem within discrete diffusion models, analogous to reinforcement learning (RL), while minimizing the KL divergence against pretrained diffusion models to preserve naturalness. To solve this RL problem, we propose a novel algorithm, DRAKES, that enables direct backpropagation of rewards through entire trajectories generated by diffusion models, by making the originally non-differentiable trajectories differentiable using the Gumbel-Softmax trick. Our theoretical analysis indicates that our approach can generate sequences that are both natural-like and yield high rewards. While similar tasks have been recently explored in diffusion models for continuous domains, our work addresses unique algorithmic and theoretical challenges specific to discrete diffusion models, which arise from their foundation in continuous-time Markov chains rather than Brownian motion. Finally, we demonstrate the effectiveness of DRAKES in generating DNA and protein sequences that optimize enhancer activity and protein stability, respectively, important tasks for gene therapies and protein-based therapeutics.

Information retrieval (IR) methods, like retrieval augmented generation, are fundamental to modern applications but often lack statistical guarantees. Conformal prediction addresses this by retrieving sets guaranteed to include relevant information, yet existing approaches produce large-sized sets, incurring high computational costs and slow response times. In this work, we introduce a score refinement method that applies a simple monotone transformation to retrieval scores, leading to significantly smaller conformal sets while maintaining their statistical guarantees. Experiments on various BEIR benchmarks validate the effectiveness of our approach in producing compact sets containing relevant information., Comment: 6 pages

The cell is arguably the most fundamental unit of life and is central to understanding biology. Accurate modeling of cells is important for this understanding as well as for determining the root causes of disease. Recent advances in artificial intelligence (AI), combined with the ability to generate large-scale experimental data, present novel opportunities to model cells. Here we propose a vision of leveraging advances in AI to construct virtual cells, high-fidelity simulations of cells and cellular systems under different conditions that are directly learned from biological data across measurements and scales. We discuss desired capabilities of such AI Virtual Cells, including generating universal representations of biological entities across scales, and facilitating interpretable in silico experiments to predict and understand their behavior using virtual instruments. We further address the challenges, opportunities and requirements to realize this vision including data needs, evaluation strategies, and community standards and engagement to ensure biological accuracy and broad utility. We envision a future where AI Virtual Cells help identify new drug targets, predict cellular responses to perturbations, as well as scale hypothesis exploration. With open science collaborations across the biomedical ecosystem that includes academia, philanthropy, and the biopharma and AI industries, a comprehensive predictive understanding of cell mechanisms and interactions has come into reach.

Recent advancements in generative models have revolutionized image generation and editing, making these tasks accessible to non-experts. This paper focuses on local image editing, particularly the task of adding new content to a loosely specified area. Existing methods often require a precise mask or a detailed description of the location, which can be cumbersome and prone to errors. We propose Click2Mask, a novel approach that simplifies the local editing process by requiring only a single point of reference (in addition to the content description). A mask is dynamically grown around this point during a Blended Latent Diffusion (BLD) process, guided by a masked CLIP-based semantic loss. Click2Mask surpasses the limitations of segmentation-based and fine-tuning dependent methods, offering a more user-friendly and contextually accurate solution. Our experiments demonstrate that Click2Mask not only minimizes user effort but also delivers competitive or superior local image manipulation results compared to SoTA methods, according to both human judgement and automatic metrics. Key contributions include the simplification of user input, the ability to freely add objects unconstrained by existing segments, and the integration potential of our dynamic mask approach within other editing methods., Comment: Project page is available at https://omeregev.github.io/click2mask/

Diffusion models excel at capturing the natural design spaces of images, molecules, DNA, RNA, and protein sequences. However, rather than merely generating designs that are natural, we often aim to optimize downstream reward functions while preserving the naturalness of these design spaces. Existing methods for achieving this goal often require ``differentiable'' proxy models (\textit{e.g.}, classifier guidance or DPS) or involve computationally expensive fine-tuning of diffusion models (\textit{e.g.}, classifier-free guidance, RL-based fine-tuning). In our work, we propose a new method to address these challenges. Our algorithm is an iterative sampling method that integrates soft value functions, which looks ahead to how intermediate noisy states lead to high rewards in the future, into the standard inference procedure of pre-trained diffusion models. Notably, our approach avoids fine-tuning generative models and eliminates the need to construct differentiable models. This enables us to (1) directly utilize non-differentiable features/reward feedback, commonly used in many scientific domains, and (2) apply our method to recent discrete diffusion models in a principled way. Finally, we demonstrate the effectiveness of our algorithm across several domains, including image generation, molecule generation, and DNA/RNA sequence generation. The code is available at \href{https://github.com/masa-ue/SVDD}{https://github.com/masa-ue/SVDD}., Comment: The code is available at https://github.com/masa-ue/SVDD

Video generation has drawn significant interest recently, pushing the development of large-scale models capable of producing realistic videos with coherent motion. Due to memory constraints, these models typically generate short video segments that are then combined into long videos. The merging process poses a significant challenge, as it requires ensuring smooth transitions and overall consistency. In this paper, we introduce Anchored Diffusion, a novel method for synthesizing relatively long and seamless videos. We extend Diffusion Transformers (DiTs) to incorporate temporal information, creating our sequence-DiT (sDiT) model for generating short video segments. Unlike previous works, we train our model on video sequences with random non-uniform temporal spacing and incorporate temporal information via external guidance, increasing flexibility and allowing it to capture both short and long-term relationships. Furthermore, during inference, we leverage the transformer architecture to modify the diffusion process, generating a batch of non-uniform sequences anchored to a common frame, ensuring consistency regardless of temporal distance. To demonstrate our method, we focus on face reenactment, the task of creating a video from a source image that replicates the facial expressions and movements from a driving video. Through comprehensive experiments, we show our approach outperforms current techniques in producing longer consistent high-quality videos while offering editing capabilities.

The pursuit of high perceptual quality in image restoration has driven the development of revolutionary generative models, capable of producing results often visually indistinguishable from real data. However, as their perceptual quality continues to improve, these models also exhibit a growing tendency to generate hallucinations - realistic-looking details that do not exist in the ground truth images. Hallucinations in these models create uncertainty about their reliability, raising major concerns about their practical application. This paper investigates this phenomenon through the lens of information theory, revealing a fundamental tradeoff between uncertainty and perception. We rigorously analyze the relationship between these two factors, proving that the global minimal uncertainty in generative models grows in tandem with perception. In particular, we define the inherent uncertainty of the restoration problem and show that attaining perfect perceptual quality entails at least twice this uncertainty. Additionally, we establish a relation between distortion, uncertainty and perception, through which we prove the aforementioned uncertainly-perception tradeoff induces the well-known perception-distortion tradeoff. We demonstrate our theoretical findings through experiments with super-resolution and inpainting algorithms. This work uncovers fundamental limitations of generative models in achieving both high perceptual quality and reliable predictions for image restoration. Thus, we aim to raise awareness among practitioners about this inherent tradeoff, empowering them to make informed decisions and potentially prioritize safety over perceptual performance.

Analyzing the cardiovascular system condition via Electrocardiography (ECG) is a common and highly effective approach, and it has been practiced and perfected over many decades. ECG sensing is non-invasive and relatively easy to acquire, and yet it is still cumbersome for holter monitoring tests that may span over hours and even days. A possible alternative in this context is Photoplethysmography (PPG): An optically-based signal that measures blood volume fluctuations, as typically sensed by conventional ``wearable devices''. While PPG presents clear advantages in acquisition, convenience, and cost-effectiveness, ECG provides more comprehensive information, allowing for a more precise detection of heart conditions. This implies that a conversion from PPG to ECG, as recently discussed in the literature, inherently involves an unavoidable level of uncertainty. In this paper we introduce a novel methodology for addressing the PPG-2-ECG conversion, and offer an enhanced classification of cardiovascular conditions using the given PPG, all while taking into account the uncertainties arising from the conversion process. We provide a mathematical justification for our proposed computational approach, and present empirical studies demonstrating its superior performance compared to state-of-the-art baseline methods.

It is now possible to conduct large scale perturbation screens with complex readout modalities, such as different molecular profiles or high content cell images. While these open the way for systematic dissection of causal cell circuits, integrated such data across screens to maximize our ability to predict circuits poses substantial computational challenges, which have not been addressed. Here, we extend two Gromov-Wasserstein Optimal Transport methods to incorporate the perturbation label for cross-modality alignment. The obtained alignment is then employed to train a predictive model that estimates cellular responses to perturbations observed with only one measurement modality. We validate our method for the tasks of cross-modality alignment and cross-modality prediction in a recent multi-modal single-cell perturbation dataset. Our approach opens the way to unified causal models of cell biology., Comment: 16 pages, 4 figures, correspondence to Aviv Regev and Romain Lopez

The origin of the transition from asymptotically reversible to asymptotically irreversible response in amorphous solids subject to oscillatory shear is still unknown. It is known that the plastic events that result from shearing always involve localized particle rearrangements, but it is unclear why some are reversible while others are not. Here we show, using simulations and models, that overlaps between particle rearrangements caused by straining the solid in alternating directions can cause the response to become irreversible when they occur frequently. As the forcing amplitude increases, plastic events become more frequent, the number of such overlaps increases, and the probability of the system returning to previous states diminishes.

The incorporation of Denoising Diffusion Models (DDMs) in the Text-to-Speech (TTS) domain is rising, providing great value in synthesizing high quality speech. Although they exhibit impressive audio quality, the extent of their semantic capabilities is unknown, and controlling their synthesized speech's vocal properties remains a challenge. Inspired by recent advances in image synthesis, we explore the latent space of frozen TTS models, which is composed of the latent bottleneck activations of the DDM's denoiser. We identify that this space contains rich semantic information, and outline several novel methods for finding semantic directions within it, both supervised and unsupervised. We then demonstrate how these enable off-the-shelf audio editing, without any further training, architectural changes or data requirements. We present evidence of the semantic and acoustic qualities of the edited audio, and provide supplemental samples: https://latent-analysis-grad-tts.github.io/speech-samples/., Comment: Accepted to ACL 2024

Early time classification algorithms aim to label a stream of features without processing the full input stream, while maintaining accuracy comparable to that achieved by applying the classifier to the entire input. In this paper, we introduce a statistical framework that can be applied to any sequential classifier, formulating a calibrated stopping rule. This data-driven rule attains finite-sample, distribution-free control of the accuracy gap between full and early-time classification. We start by presenting a novel method that builds on the Learn-then-Test calibration framework to control this gap marginally, on average over i.i.d. instances. As this algorithm tends to yield an excessively high accuracy gap for early halt times, our main contribution is the proposal of a framework that controls a stronger notion of error, where the accuracy gap is controlled conditionally on the accumulated halt times. Numerical experiments demonstrate the effectiveness, applicability, and usefulness of our method. We show that our proposed early stopping mechanism reduces up to 94% of timesteps used for classification while achieving rigorous accuracy gap control.

Deep Generative Models (DGMs) are versatile tools for learning data representations while adequately incorporating domain knowledge such as the specification of conditional probability distributions. Recently proposed DGMs tackle the important task of comparing data sets from different sources. One such example is the setting of contrastive analysis that focuses on describing patterns that are enriched in a target data set compared to a background data set. The practical deployment of those models often assumes that DGMs naturally infer interpretable and modular latent representations, which is known to be an issue in practice. Consequently, existing methods often rely on ad-hoc regularization schemes, although without any theoretical grounding. Here, we propose a theory of identifiability for comparative DGMs by extending recent advances in the field of non-linear independent component analysis. We show that, while these models lack identifiability across a general class of mixing functions, they surprisingly become identifiable when the mixing function is piece-wise affine (e.g., parameterized by a ReLU neural network). We also investigate the impact of model misspecification, and empirically show that previously proposed regularization techniques for fitting comparative DGMs help with identifiability when the number of latent variables is not known in advance. Finally, we introduce a novel methodology for fitting comparative DGMs that improves the treatment of multiple data sources via multi-objective optimization and that helps adjust the hyperparameter for the regularization in an interpretable manner, using constrained optimization. We empirically validate our theory and new methodology using simulated data as well as a recent data set of genetic perturbations in cells profiled via single-cell RNA sequencing., Comment: 45 pages, 3 figures

Linear solvers are major computational bottlenecks in a wide range of decision support and optimization computations. The challenges become even more pronounced on heterogeneous hardware, where traditional sparse numerical linear algebra methods are often inefficient. For example, methods for solving ill-conditioned linear systems have relied on conditional branching, which degrades performance on hardware accelerators such as graphical processing units (GPUs). To improve the efficiency of solving ill-conditioned systems, our computational strategy separates computations that are efficient on GPUs from those that need to run on traditional central processing units (CPUs). Our strategy maximizes the reuse of expensive CPU computations. Iterative methods, which thus far have not been broadly used for ill-conditioned linear systems, play an important role in our approach. In particular, we extend ideas from [1] to implement iterative refinement using inexact LU factors and flexible generalized minimal residual (FGMRES), with the aim of efficient performance on GPUs. We focus on solutions that are effective within broader application contexts, and discuss how early performance tests could be improved to be more predictive of the performance in a realistic environment, Comment: 15 pages, 8 figures, 5 tables

Our main result is a reduction from worst-case lattice problems such as GapSVP and SIVP to a certain learning problem. This learning problem is a natural extension of the `learning from parity with error' problem to higher moduli. It can also be viewed as the problem of decoding from a random linear code. This, we believe, gives a strong indication that these problems are hard. Our reduction, however, is quantum. Hence, an efficient solution to the learning problem implies a quantum algorithm for GapSVP and SIVP. A main open question is whether this reduction can be made classical (i.e., non-quantum). We also present a (classical) public-key cryptosystem whose security is based on the hardness of the learning problem. By the main result, its security is also based on the worst-case quantum hardness of GapSVP and SIVP. The new cryptosystem is much more efficient than previous lattice-based cryptosystems: the public key is of size $\tilde{O}(n^2)$ and encrypting a message increases its size by a factor of $\tilde{O}(n)$ (in previous cryptosystems these values are $\tilde{O}(n^4)$ and $\tilde{O}(n^2)$, respectively). In fact, under the assumption that all parties share a random bit string of length $\tilde{O}(n^2)$, the size of the public key can be reduced to $\tilde{O}(n)$., Comment: Identical to the version posted in 2009, apart from fixing a bug in the proof of Claim 3.13

Training on multiple modalities of input can augment the capabilities of a language model. Here, we ask whether such a training regime can improve the quality and efficiency of these systems as well. We focus on text--audio and introduce Whisbert, which is inspired by the text--image approach of FLAVA (Singh et al., 2022). In accordance with Babylm guidelines (Warstadt et al., 2023), we pretrain Whisbert on a dataset comprising only 100 million words plus their corresponding speech from the word-aligned version of the People's Speech dataset (Galvez et al., 2021). To assess the impact of multimodality, we compare versions of the model that are trained on text only and on both audio and text simultaneously. We find that while Whisbert is able to perform well on multimodal masked modeling and surpasses the Babylm baselines in most benchmark tasks, it struggles to optimize its complex objective and outperform its text-only Whisbert baseline., Comment: Published at the BabyLM Challenge, a shared task co-sponsored by CMCL 2023 and CoNLL 2023, hosted by EMNLP 2023

Let $K$ be a convex body in $\mathbb{R}^n$, let $L$ be a lattice with covolume one, and let $\eta>0$. We say that $K$ and $L$ form an $\eta$-smooth cover if each point $x \in \mathbb{R}^n$ is covered by $(1 \pm \eta) vol(K)$ translates of $K$ by $L$. We prove that for any positive $\sigma, \eta$, asymptotically as $n \to \infty$, for any $K$ of volume $n^{3+\sigma}$, one can find a lattice $L$ for which $L, K$ form an $\eta$-smooth cover. Moreover, this property is satisfied with high probability for a lattice chosen randomly, according to the Haar-Siegel measure on the space of lattices. Similar results hold for random construction A lattices, albeit with a worse power law, provided the ratio between the covering and packing radii of $\mathbb{Z}^n$ with respect to $K$ is at most polynomial in $n$. Our proofs rely on a recent breakthrough by Dhar and Dvir on the discrete Kakeya problem.

A key element of computer-assisted surgery systems is phase recognition of surgical videos. Existing phase recognition algorithms require frame-wise annotation of a large number of videos, which is time and money consuming. In this work we join concepts of graph segmentation with self-supervised learning to derive a random-walk solution for per-frame phase prediction. Furthermore, we utilize within our method two forms of weak supervision: sparse timestamps or few-shot learning. The proposed algorithm enjoys low complexity and can operate in lowdata regimes. We validate our method by running experiments with the public Cholec80 dataset of laparoscopic cholecystectomy videos, demonstrating promising performance in multiple setups.

The discovery that memory of particle configurations and plastic events can be stored in amorphous solids subject to oscillatory shear has spurred research into methods for storing and retrieving information from these materials. However, it is unclear to what extent the ability to store memory is affected by thermal fluctuations and other environmental noises, which are expected to be relevant in realistic situations. Here, we show that while memory has a long lifetime at low temperatures, thermal fluctuations eventually lead to a catastrophic loss of memory, resulting in the erasure of most or all of the stored information within a few forcing cycles. We observe that an escape from the memory-retaining state (limit cycle) is triggered by a change in the switching of plastic events, leading to a cascade of new plastic events that were not present in the original limit cycle. The displacements from the new plastic events change the particle configuration which leads to the loss of memory. We further show that the rate of escaping from a limit cycle increases in a non-Arrhenius manner as a function of temperature, and the probability of staying in a limit cycle decays exponentially with an increase in the shearing frequency. These results have important implications for memory storage since increasing the temperature offers a means of effectively erasing existing memories and allowing for the imprinting of new ones that can then be stored for a long time at low temperatures., Comment: 7 pages, 9 figures

Processing-in-memory (PIM) is attractive to overcome the limitations of modern computing systems. Numerous PIM systems exist, varying by the technologies and logic techniques used. Successful operation of specific logic functions is crucial for effective processing-in-memory. Memristive non-stateful logic techniques are compatible with CMOS logic and can be integrated into a 1T1R memory array, similar to commercial RRAM products. This paper analyzes and demonstrates two non-stateful logic techniques: 1T1R logic and scouting logic. As a first step, the used 1T1R SiO\textsubscript{x} valence change mechanism memristors are characterized in reference to their feasibility to perform logic functions. Various logical functions of the two logic techniques are experimentally demonstrated, showing correct functionality in all cases. Following the results, the challenges and limitations of the RRAM characteristics and 1T1R configuration for the application in logical functions are discussed., Comment: 5 pages, 6 figures

Self-supervised learning (SSL) has led to important breakthroughs in computer vision by allowing learning from large amounts of unlabeled data. As such, it might have a pivotal role to play in biomedicine where annotating data requires a highly specialized expertise. Yet, there are many healthcare domains for which SSL has not been extensively explored. One such domain is endoscopy, minimally invasive procedures which are commonly used to detect and treat infections, chronic inflammatory diseases or cancer. In this work, we study the use of a leading SSL framework, namely Masked Siamese Networks (MSNs), for endoscopic video analysis such as colonoscopy and laparoscopy. To fully exploit the power of SSL, we create sizable unlabeled endoscopic video datasets for training MSNs. These strong image representations serve as a foundation for secondary training with limited annotated datasets, resulting in state-of-the-art performance in endoscopic benchmarks like surgical phase recognition during laparoscopy and colonoscopic polyp characterization. Additionally, we achieve a 50% reduction in annotated data size without sacrificing performance. Thus, our work provides evidence that SSL can dramatically reduce the need of annotated data in endoscopy., Comment: Accepted to MICCAI 2023

We show that $n$-bit integers can be factorized by independently running a quantum circuit with $\tilde{O}(n^{3/2})$ gates for $\sqrt{n}+4$ times, and then using polynomial-time classical post-processing. The correctness of the algorithm relies on a number-theoretic heuristic assumption reminiscent of those used in subexponential classical factorization algorithms. It is currently not clear if the algorithm can lead to improved physical implementations in practice.

We prove that for any integral lattice $\mathcal{L} \subset \mathbb{R}^n$ (that is, a lattice $\mathcal{L}$ such that the inner product $\langle \mathbf{y}_1,\mathbf{y}_2 \rangle$ is an integer for all $\mathbf{y}_1, \mathbf{y}_2 \in \mathcal{L}$) and any positive integer $k$, \[ |\{ \mathbf{y} \in \mathcal{L} \ : \ \|\mathbf{y}\|^2 = k\}| \leq 2 \binom{n+2k-2}{2k-1} \; , \] giving a nearly tight reverse Minkowski theorem for integral lattices.

This paper presents an analysis of the fundamental limits on energy efficiency in both digital and analog in-memory computing architectures, and compares their performance to single instruction, single data (scalar) machines specifically in the context of machine inference. The focus of the analysis is on how efficiency scales with the size, arithmetic intensity, and bit precision of the computation to be performed. It is shown that analog, in-memory computing architectures can approach arbitrarily high energy efficiency as both the problem size and processor size scales., Comment: 17 pages, 10 figures