Search

The coupling between evolutionary and ecological changes (eco-evolutionary dynamics) has been shown to be relevant among diverse species, and is also of interest outside of ecology, i.e. in cancer evolution. These dynamics play an important role in determining survival in response to climate change, motivating the need for mathematical models to capture this often complex interplay. Population genetics models incorporating eco-evolutionary dynamics often sacrifice analytical tractability to model the complexity of real systems, and do not explicitly consider the effect of population heterogeneity. In order to allow for population heterogeneity, transient, and long-term dynamics, while retaining tractability, we generalise a moment-based method applicable in the regime of small segregational variance to the case of time-dependent mortality and birth. These results are applied to a predator-prey model, where ecological parameters such as the contact rate between species are trait-structured. The trait-distribution of the prey species is shown to be approximately Gaussian with constant variance centered on the mean trait, which is asymptotically governed by an autonomous ODE. In this way, we make explicit the impact of eco-evolutionary dynamics on the transient behaviour and long-term fate of the prey species.

Despite its high prevalence and impact on the lives of those affected, a complete understanding of the cause of inflammatory bowel disease (IBD) is lacking. In this paper, we investigate a novel mechanism which proposes that mutant epithelial cells are significant to the progression of IBD since they promote inflammation and are resistant to death. We develop a simple model encapsulating the propagation of mutant epithelial cells and immune cells which results from interactions with the intestinal barrier and bacteria. Motivated by the slow growth of mutant epithelial cells, and relatively slow response rate of the adaptive immune system, we combine geometric singular perturbation theory with matched asymptotic expansions to determine the one-dimensional slow invariant manifold that characterises the leading order dynamics at all times beyond a passive initial adjustment phase. The dynamics on this manifold are controlled by a bifurcation parameter, $\sigma$, which depends on the ratio of growth to decay rates of all components except mutants and determines three distinct classes of permanent-form travelling waves that describe the propagation of mutant epithelial and immune cells. These are obtained from scalar reaction-diffusion equations with the reaction being (i) a bistable nonlinearity with a cut-off, (ii) a cubic Fisher nonlinearity and (iii) a KPP or Fisher nonlinearity. Our results suggest that mutant epithelial cells are critical to the progression of IBD. However, their effect on the speed of progression is subdominant., Comment: corrected formatting in the abstract. Manuscript unchanged

Bias significantly undermines both the accuracy and trustworthiness of machine learning models. To date, one of the strongest biases observed in image classification models is texture bias-where models overly rely on texture information rather than shape information. Yet, existing approaches for measuring and mitigating texture bias have not been able to capture how textures impact model robustness in real-world settings. In this work, we introduce the Texture Association Value (TAV), a novel metric that quantifies how strongly models rely on the presence of specific textures when classifying objects. Leveraging TAV, we demonstrate that model accuracy and robustness are heavily influenced by texture. Our results show that texture bias explains the existence of natural adversarial examples, where over 90% of these samples contain textures that are misaligned with the learned texture of their true label, resulting in confident mispredictions., Comment: Accepted to IEEE Secure and Trustworthy Machine Learning (SaTML)

The widespread use of Large Language Models (LLMs) in society creates new information security challenges for developers, organizations, and end-users alike. LLMs are trained on large volumes of data, and their susceptibility to reveal the exact contents of the source training datasets poses security and safety risks. Although current alignment procedures restrict common risky behaviors, they do not completely prevent LLMs from leaking data. Prior work demonstrated that LLMs may be tricked into divulging training data by using out-of-distribution queries or adversarial techniques. In this paper, we demonstrate a simple, query-based decompositional method to extract news articles from two frontier LLMs. We use instruction decomposition techniques to incrementally extract fragments of training data. Out of 3723 New York Times articles, we extract at least one verbatim sentence from 73 articles, and over 20% of verbatim sentences from 6 articles. Our analysis demonstrates that this method successfully induces the LLM to generate texts that are reliable reproductions of news articles, meaning that they likely originate from the source training dataset. This method is simple, generalizable, and does not fine-tune or change the production model. If replicable at scale, this training data extraction methodology could expose new LLM security and safety vulnerabilities, including privacy risks and unauthorized data leaks. These implications require careful consideration from model development to its end-use.

Machine-learning (ML) models in flow cytometry have the potential to reduce error rates, increase reproducibility, and boost the efficiency of clinical labs. While numerous ML models for flow cytometry data have been proposed, few studies have described the clinical deployment of such models. Realizing the potential gains of ML models in clinical labs requires not only an accurate model, but infrastructure for automated inference, error detection, analytics and monitoring, and structured data extraction. Here, we describe an ML model for detection of Acute Myeloid Leukemia (AML), along with the infrastructure supporting clinical implementation. Our infrastructure leverages the resilience and scalability of the cloud for model inference, a Kubernetes-based workflow system that provides model reproducibility and resource management, and a system for extracting structured diagnoses from full-text reports. We also describe our model monitoring and visualization platform, an essential element for ensuring continued model accuracy. Finally, we present a post-deployment analysis of impacts on turn-around time and compare production accuracy to the original validation statistics.

The influence of textures on machine learning models has been an ongoing investigation, specifically in texture bias/learning, interpretability, and robustness. However, due to the lack of large and diverse texture data available, the findings in these works have been limited, as more comprehensive evaluations have not been feasible. Image generative models are able to provide data creation at scale, but utilizing these models for texture synthesis has been unexplored and poses additional challenges both in creating accurate texture images and validating those images. In this work, we introduce an extensible methodology and corresponding new dataset for generating high-quality, diverse texture images capable of supporting a broad set of texture-based tasks. Our pipeline consists of: (1) developing prompts from a range of descriptors to serve as input to text-to-image models, (2) adopting and adapting Stable Diffusion pipelines to generate and filter the corresponding images, and (3) further filtering down to the highest quality images. Through this, we create the Prompted Textures Dataset (PTD), a dataset of 362,880 texture images that span 56 textures. During the process of generating images, we find that NSFW safety filters in image generation pipelines are highly sensitive to texture (and flag up to 60\% of our texture images), uncovering a potential bias in these models and presenting unique challenges when working with texture data. Through both standard metrics and a human evaluation, we find that our dataset is high quality and diverse.

The Tor anonymity network allows users such as political activists and those under repressive governments to protect their privacy when communicating over the internet. At the same time, Tor has been demonstrated to be vulnerable to several classes of deanonymizing attacks that expose user behavior and identities. Prior work has shown that these threats can be mitigated by leveraging trusted execution environments (TEEs). However, previous proposals assume that all relays in the network will be TEE-based-which as a practical matter is unrealistic. In this work, we introduce ParTEETor, a Tor-variant system, which leverages partial deployments of TEEs to thwart known attacks. We study two modes of operation: non-policy and policy. Non-policy mode uses the existing Tor relay selection algorithm to provide users incident security. Policy mode extends the relay selection algorithm to address the classes of attacks by enforcing a specific TEE circuit configuration. We evaluate ParTEETor for security, performance, and privacy. Our evaluation demonstrates that at even a small TEE penetration (e.g., 10% of relays are TEE-based), users can reach performance of Tor today while enforcing a security policy to guarantee protection from at least two classes of attacks. Overall, we find that partial deployments of TEEs can substantially improve the security of Tor, without a significant impact on performance or privacy.

The sheer scale and diversity of transportation make it a formidable sector to decarbonize. Here, we consider an emerging opportunity to reduce carbon emissions: the growing adoption of semi-autonomous vehicles, which can be programmed to mitigate stop-and-go traffic through intelligent speed commands and, thus, reduce emissions. But would such dynamic eco-driving move the needle on climate change? A comprehensive impact analysis has been out of reach due to the vast array of traffic scenarios and the complexity of vehicle emissions. We address this challenge with large-scale scenario modeling efforts and by using multi-task deep reinforcement learning with a carefully designed network decomposition strategy. We perform an in-depth prospective impact assessment of dynamic eco-driving at 6,011 signalized intersections across three major US metropolitan cities, simulating a million traffic scenarios. Overall, we find that vehicle trajectories optimized for emissions can cut city-wide intersection carbon emissions by 11-22%, without harming throughput or safety, and with reasonable assumptions, equivalent to the national emissions of Israel and Nigeria, respectively. We find that 10% eco-driving adoption yields 25%-50% of the total reduction, and nearly 70% of the benefits come from 20% of intersections, suggesting near-term implementation pathways. However, the composition of this high-impact subset of intersections varies considerably across different adoption levels, with minimal overlap, calling for careful strategic planning for eco-driving deployments. Moreover, the impact of eco-driving, when considered jointly with projections of vehicle electrification and hybrid vehicle adoption remains significant. More broadly, this work paves the way for large-scale analysis of traffic externalities, such as time, safety, and air quality, and the potential impact of solution strategies., Comment: In review

Let ${\mathfrak g}$ be a Garding-Dirichlet operator on the set S(n) of symmetric $n\times n$ matrices. We assume that ${\mathfrak g}$ is $I$-central, that is, $D_I {\mathfrak g} = k I$ for some $k>0$. Then $$ {\mathfrak g}(A)^{1\over N} \ \geq\ {\mathfrak g}(I)^{1\over N} (\det\, A)^{1\over n} \qquad \forall\, A>0. $$ From work of Guo, Phong, Tong, Abja, Dinew, Olive and many others, this inequality has important applications.

The $\beta$ Pictoris system is the closest known stellar system with directly detected gas giant planets, an edge-on circumstellar disc, and evidence of falling sublimating bodies and transiting exocomets. The inner planet, $\beta$ Pictoris c, has also been indirectly detected with radial velocity (RV) measurements. The star is a known $\delta$ Scuti pulsator, and the long-term stability of these pulsations opens up the possibility of indirectly detecting the gas giant planets through time delays of the pulsations due to a varying light travel time. We search for phase shifts in the $\delta$ Scuti pulsations consistent with the known planets $\beta$ Pictoris b and c and carry out an analysis of the stellar pulsations of $\beta$ Pictoris over a multi-year timescale. We used photometric data collected by the BRITE-Constellation, bRing, ASTEP, and TESS to derive a list of the strongest and most significant $\delta$ Scuti pulsations. We carried out an analysis with the open-source python package maelstrom to study the stability of the pulsation modes of $\beta$ Pictoris in order to determine the long-term trends in the observed pulsations. We did not detect the expected signal for $\beta$ Pictoris b or $\beta$ Pictoris c. The expected time delay is 6 seconds for $\beta$ Pictoris c and 24 seconds for $\beta$ Pictoris b. With simulations, we determined that the photometric noise in all the combined data sets cannot reach the sensitivity needed to detect the expected timing drifts. An analysis of the pulsational modes of $\beta$ Pictoris using maelstrom showed that the modes themselves drift on the timescale of a year, fundamentally limiting our ability to detect exoplanets around $\beta$ Pictoris via pulsation timing., Comment: 16 pages, 16 figures, 4 tables, accepted for publication in A&A

We introduce Geometric Neural Operators (GNPs) for accounting for geometric contributions in data-driven deep learning of operators. We show how GNPs can be used (i) to estimate geometric properties, such as the metric and curvatures, (ii) to approximate Partial Differential Equations (PDEs) on manifolds, (iii) learn solution maps for Laplace-Beltrami (LB) operators, and (iv) to solve Bayesian inverse problems for identifying manifold shapes. The methods allow for handling geometries of general shape including point-cloud representations. The developed GNPs provide approaches for incorporating the roles of geometry in data-driven learning of operators.

In this work, we investigate \textit{texture learning}: the identification of textures learned by object classification models, and the extent to which they rely on these textures. We build texture-object associations that uncover new insights about the relationships between texture and object classes in CNNs and find three classes of results: associations that are strong and expected, strong and not expected, and expected but not present. Our analysis demonstrates that investigations in texture learning enable new methods for interpretability and have the potential to uncover unexpected biases., Comment: Accepted to ICLR 2024, Tiny Papers Track

Computing reduced-order models using non-intrusive methods is particularly attractive for systems that are simulated using black-box solvers. However, obtaining accurate data-driven models can be challenging, especially if the underlying systems exhibit large-amplitude transient growth. Although these systems may evolve near a low-dimensional subspace that can be easily identified using standard techniques such as Proper Orthogonal Decomposition (POD), computing accurate models often requires projecting the state onto this subspace via a non-orthogonal projection. While appropriate oblique projection operators can be computed using intrusive techniques that leverage the form of the underlying governing equations, purely data-driven methods currently tend to achieve dimensionality reduction via orthogonal projections, and this can lead to models with poor predictive accuracy. In this paper, we address this issue by introducing a non-intrusive framework designed to simultaneously identify oblique projection operators and reduced-order dynamics. In particular, given training trajectories and assuming reduced-order dynamics of polynomial form, we fit a reduced-order model by solving an optimization problem over the product manifold of a Grassmann manifold, a Stiefel manifold, and several linear spaces (as many as the tensors that define the low-order dynamics). Furthermore, we show that the gradient of the cost function with respect to the optimization parameters can be conveniently written in closed-form, so that there is no need for automatic differentiation. We compare our formulation with state-of-the-art methods on three examples: a three-dimensional system of ordinary differential equations, the complex Ginzburg-Landau (CGL) equation, and a two-dimensional lid-driven cavity flow at Reynolds number Re = 8300.

Conversational question answering (convQA) over knowledge graphs (KGs) involves answering multi-turn natural language questions about information contained in a KG. State-of-the-art methods of ConvQA often struggle with inexplicit question-answer pairs. These inputs are easy for human beings to understand given a conversation history, but hard for a machine to interpret, which can degrade ConvQA performance. To address this problem, we propose a reinforcement learning (RL) based model, CornNet, which utilizes question reformulations generated by large language models (LLMs) to improve ConvQA performance. CornNet adopts a teacher-student architecture where a teacher model learns question representations using human writing reformulations, and a student model to mimic the teacher model's output via reformulations generated by LLMs. The learned question representation is then used by an RL model to locate the correct answer in a KG. Extensive experimental results show that CornNet outperforms state-of-the-art convQA models.

While Large Language Models (LLMs) display versatile functionality, they continue to generate harmful, biased, and toxic content, as demonstrated by the prevalence of human-designed jailbreaks. In this work, we present Tree of Attacks with Pruning (TAP), an automated method for generating jailbreaks that only requires black-box access to the target LLM. TAP utilizes an attacker LLM to iteratively refine candidate (attack) prompts until one of the refined prompts jailbreaks the target. In addition, before sending prompts to the target, TAP assesses them and prunes the ones unlikely to result in jailbreaks, reducing the number of queries sent to the target LLM. In empirical evaluations, we observe that TAP generates prompts that jailbreak state-of-the-art LLMs (including GPT4-Turbo and GPT4o) for more than 80% of the prompts. This significantly improves upon the previous state-of-the-art black-box methods for generating jailbreaks while using a smaller number of queries than them. Furthermore, TAP is also capable of jailbreaking LLMs protected by state-of-the-art guardrails, e.g., LlamaGuard., Comment: Accepted for presentation at NeurIPS 2024. Code: https://github.com/RICommunity/TAP