Search for: All records

Biomedical knowledge graphs (KGs) encode rich, structured information critical for drug discovery tasks, but extracting meaningful insights from large-scale KGs remains challenging due to their complex structure. Existing biomedical subgraph retrieval methods are tailored for graph neural networks (GNNs), limiting compatibility with other paradigms, including large language models (LLMs). We introduce K-Paths, a model-agnostic retrieval framework that extracts structured, diverse, and biologically meaningful multi-hop paths from dense biomedical KGs. These paths enable prediction of unobserved drug-drug and drug-disease interactions, including those involving entities not seen during training, thus supporting inductive reasoning. K-Paths is training-free and employs a diversity-aware adaptation of Yen's algorithm to extract the K shortest loopless paths between entities in a query, prioritizing biologically relevant and relationally diverse connections. These paths serve as concise, interpretable reasoning chains that can be directly integrated with LLMs or GNNs to improve generalization, accuracy, and enable explainable inference. Experiments on benchmark datasets show that K-Paths improves zero-shot reasoning across state-of-the-art LLMs. For instance, Tx-Gemma 27B improves by 19.8 and 4.0 F1 points on interaction severity prediction and drug repurposing tasks, respectively. Llama 70B achieves gains of 8.5 and 6.2 points on the same tasks. K-Paths also boosts the training efficiency of EmerGNN, a state-of-the-art GNN, by reducing the KG size by 90% while maintaining predictive performance. Beyond efficiency, K-Paths bridges the gap between KGs and LLMs, enabling scalable and explainable LLM-augmented scientific discovery. We release our code and the retrieved paths as a benchmark for inductive reasoning. 

Traditional robotic vehicle control algorithms, implemented on digital devices with firmware, result in high power consumption and system complexity. Advanced control systems based on different device physics are essential for the advancement of sophisticated robotic vehicles and miniature mobile robots. Here, we present a nanoelectronics-enabled analog control system mimicking conventional controllers’ dynamic responses for real-time robotic controls, substantially reducing training cost, power consumption, and footprint. This system uses a reservoir computing network with interconnected memristive channels made from layered semiconductors. The network’s nonlinear switching and short-term memory characteristics effectively map input sensory signals to high-dimensional data spaces, enabling the generation of motor control signals with a simply trained readout layer. This approach minimizes software and analog-to-digital conversions, enhancing energy and resource efficiency. We demonstrate this system with two control tasks: rover target tracking and drone lever balancing, achieving similar performance to traditional controllers with ~10-microwatt power consumption. This work paves the way for ultralow-power edge computing in miniature robotic systems. 

We study a variant of the contextual bandit problem where an agent can intervene through a set of stochastic expert policies. Given a fixed context, each expert samples actions from a fixed conditional distribution. The agent seeks to remain competitive with the “best” among the given set of experts. We propose the Divergence-based Upper Confidence Bound (D-UCB) algorithm that uses importance sampling to share information across experts and provide horizon-independent constant regret bounds that only scale linearly in the number of experts. We also provide the Empirical D-UCB (ED-UCB) algorithm that can function with only approximate knowledge of expert distributions. Further, we investigate the episodic setting where the agent interacts with an environment that changes over episodes. Each episode can have different context and reward distributions resulting in the best expert changing across episodes. We show that by bootstrapping from\(\mathcal {O}(N\log (NT^2\sqrt {E}))\)samples, ED-UCB guarantees a regret that scales as\(\mathcal {O}(E(N+1) + \frac{N\sqrt {E}}{T^2})\)forNexperts overEepisodes, each of lengthT. We finally empirically validate our findings through simulations. 

Abstract An emerging paradigm in modern electronics is that of CMOS+$${\mathsf{X}}$$ $X$requiring the integration of standard CMOS technology with novel materials and technologies denoted by$${\mathsf{X}}$$ $X$. In this context, a crucial challenge is to develop accurate circuit models for$${\mathsf{X}}$$ $X$that are compatible with standard models for CMOS-based circuits and systems. In this perspective, we present physics-based, experimentally benchmarked modular circuit models that can be used to evaluate a class of CMOS+$${\mathsf{X}}$$ $X$systems, where$${\mathsf{X}}$$ $X$denotes magnetic and spintronic materials and phenomena. This class of materials is particularly challenging because they go beyond conventional charge-based phenomena and involve the spin degree of freedom which involves non-trivial quantum effects. Starting from density matrices—the central quantity in quantum transport—using well-defined approximations, it is possible to obtain spin-circuits that generalize ordinary circuit theory to 4-component currents and voltages (1 for charge and 3 for spin). With step-by-step examples that progressively become more complex, we illustrate how the spin-circuit approach can be used to start from the physics of magnetism and spintronics to enable accurate system-level evaluations. We believe the core approach can be extended to include other quantum degrees of freedom like valley and pseudospins starting from corresponding density matrices. 

This paper reports on the culmination of an NSF Scholarships in Science, Technology, Engineering and Mathematics (S-STEM) awarded to a two-year college located in a metro area with high rates of concentrated poverty and low levels of educational attainment. This two-year college is a minority-serving institution with curriculum to prepare students majoring in engineering to transfer and complete a baccalaureate degree at a four-year university. The Engineering Scholars Program (ESP) was established in fall 2019 to award students majoring in engineering annual scholarships of up to $6000, depending on financial need. In addition to supporting students through scholarships, the program engages scholars in professional development activities inclusive of academic seminars, extracurricular events, and undergraduate research opportunities in collaboration with the local four-year university. The program also established a mentorship structure with faculty mentors, student peer mentors, and academic advising. In addition to supporting scholars at the two-year college, the ESP provides support for a portion of cohorts that have transferred to the local four-year university and remained connected to the program. To date, the ESP has awarded a total of 131 semester long scholarships; 16 in year one (2019-2020), 28 in year two (2020-2021), 35 in year three (2021-2022), including six transfers, 38 in year four (2022-2023), including eight transfers, and 28 in year five (2023-2024), including 10 transfers. In year three, the ESP was awarded supplemental funding to support a larger portion of students and transfer cohorts; this helped reduce the financial burdens resulting from exacerbated financial needs due to the COVID-19 pandemic during years two and three of this project. This paper details the progress made towards the achievement of the program goals of creating a welcoming STEM climate at the two-year college, increasing the participation and persistence in engineering among economically disadvantaged students, and establishing transfer support to the local four-year university. Program evaluation findings have identified several opportunities for sustaining scholar transfer support outside of the financial support provided in the form of scholarships. These opportunities fell into two major themes: (1) peer-led transfer support inclusive of connecting transferred students and students preparing for transfer with emphasis on navigating different university structures, and (2) collaboration across engineering disciplines to develop and offer interdisciplinary undergraduate research and/or collaborative work on other projects. Furthermore, research findings from interviews with scholars provided additional context for taking action on program outcomes while also enhancing the understanding of how participation in a collaborative cohort experience can contribute to students’ membership within the STEM community and the construction of their own STEM identity. Although formal financial support sunsets during the final year of the ESP, program and research findings have identified programmatic elements that provide key support for students and can be sustained into the future. This paper reports on the program strategy for meeting the future needs of scholars at both the two-year college and the four-year transfer university. 

Given starting and ending positions and velocities, L2 bounds on the acceleration and velocity, and the restriction to no more than two constant control inputs, this paper provides routines to compute the minimal-time path. Closed form solutions are provided for reaching a position in minimum time with and without a velocity bound, and for stopping at the goal position. A numeric solver is used to reach a goal position and velocity with no more than two constant control inputs. If a cruising phase at the terminal velocity is needed, this requires solving a non-linear equation with a single parameter. Code is provided on GitHub 1 , extended paper version at [1]. [1] https://github.com/RoboticSwarmControl/MinTimeL2pathsConstraints/ 

In this work, we propose a predator-prey system with a Holling type Ⅱ functional response and study its dynamics when the prey exhibits vigilance behavior to avoid predation and predators exhibit cooperative hunting. We provide conditions for existence and the local and global stability of equilibria. We carry out detailed bifurcation analysis and find the system to experience Hopf, saddle-node, and transcritical bifurcations. Our results show that increased prey vigilance can stabilize the system, but when vigilance levels are too high, it causes a decrease in the population density of prey and leads to extinction. When hunting cooperation is intensive, it can destabilize the system, and can also induce bi-stability phenomenon. Furthermore, it can reduce the population density of both prey and predators and also change the stability of a coexistence state. We provide numerical experiments to validate our theoretical results and discuss ecological implications. 

Large-scale neural network models combining text and images have made incredible progress in recent years. However, it remains an open question to what extent such models encode compositional representations of the concepts over which they operate, such as correctly identifying red cube by reasoning over the constituents red and cube. In this work, we focus on the ability of a large pretrained vision and language model (CLIP) to encode compositional concepts and to bind variables in a structure-sensitive way (e.g., differentiating cube behind sphere from sphere behind cube). To inspect the performance of CLIP, we compare several architectures from research on compositional distributional semantics models (CDSMs), a line of research that attempts to implement traditional compositional linguistic structures within embedding spaces. We benchmark them on three synthetic datasets– singleobject, two-object, and relational– designed to test concept binding. We find that CLIP can compose concepts in a single-object setting, but in situations where concept binding is needed, performance drops dramatically. At the same time, CDSMs also perform poorly, with best performance at chance level. 

Ever since human society entered the age of social media, every user has had a considerable amount of visual content stored online and shared in variant virtual communities. As an efficient information circulation measure, disastrous consequences are possible if the contents of images are tampered with by malicious actors. Specifically, we are witnessing the rapid development of machine learning (ML) based tools like DeepFake apps. They are capable of exploiting images on social media platforms to mimic a potential victim without their knowledge or consent. These content manipulation attacks can lead to the rapid spread of misinformation that may not only mislead friends or family members but also has the potential to cause chaos in public domains. Therefore, robust image authentication is critical to detect and filter off manipulated images. In this paper, we introduce a system that accurately AUthenticates SOcial MEdia images (AUSOME) uploaded to online platforms leveraging spectral analysis and ML. Images from DALL-E 2 are compared with genuine images from the Stanford image dataset. Discrete Fourier Transform (DFT) and Discrete Cosine Transform (DCT) are used to perform a spectral comparison. Additionally, based on the differences in their frequency response, an ML model is proposed to classify social media images as genuine or AI-generated. Using real-world scenarios, the AUSOME system is evaluated on its detection accuracy. The experimental results are encouraging and they verified the potential of the AUSOME scheme in social media image authentications.