A5049062714- Reinforcement Learning in Robotics
- Advanced Bandit Algorithms Research
- Topic Modeling
- Data Stream Mining Techniques
- Natural Language Processing Techniques
- Advanced biosensing and bioanalysis techniques
- Adversarial Robustness in Machine Learning
- Physics of Superconductivity and Magnetism
- Quantum and electron transport phenomena
- Topological Materials and Phenomena
- Surface and Thin Film Phenomena
- Advanced Fluorescence Microscopy Techniques
- Explainable Artificial Intelligence (XAI)
- Graphene and Nanomaterials Applications
- Advancements in Semiconductor Devices and Circuit Design
- Speech Recognition and Synthesis
- Optimization and Search Problems
- Nanowire Synthesis and Applications
- Evolutionary Algorithms and Applications
- Magnetic properties of thin films
- Machine Learning and Algorithms
- COVID-19 Impact on Reproduction
- Wind Energy Research and Development
- Icing and De-icing Technologies
- Modular Robots and Swarm Intelligence
University of Minnesota
2024-2025
University of Minnesota System
2024
Twin Cities Orthopedics
2024
Google (United States)
2023
Bar-Ilan University
2016-2023
Technion – Israel Institute of Technology
2008-2022
The study of graphene-based antivirals is still at a nascent stage and the photothermal antiviral properties graphene have yet to be studied. Here, we design synthesize sulfonated magnetic nanoparticles functionalized with reduced oxide (SMRGO) capture photothermally destroy herpes simplex virus type 1 (HSV-1). Graphene sheets were uniformly anchored spherical (MNPs) varying size between ∼5 25 nm. Fourier-transform infrared spectroscopy (FT-IR) confirmed sulfonation anchoring MNPs on sheets....
Trust region policy optimization (TRPO) is a popular and empirically successful search algorithm in Reinforcement Learning (RL) which surrogate problem, that restricts consecutive policies to be ‘close’ one another, iteratively solved. Nevertheless, TRPO has been considered heuristic inspired by Conservative Policy Iteration (CPI). We show the adaptive scaling mechanism used fact natural “RL version” of traditional trust-region methods from convex analysis. first analyze planning setting, we...
Abstract Studies of nanoscale superconducting structures have revealed various physical phenomena and led to the development a wide range applications. Most these studies concentrated on one- two-dimensional due lack approaches for creation fully engineered three-dimensional (3D) nanostructures. Here, we present ‘bottom-up’ method create 3D nanostructures with prescribed multiscale organization using DNA-based self-assembly methods. We assemble DNA superlattices from octahedral frames...
DNA-based self-assembly methods have demonstrated powerful and unique capabilities to encode nanomaterial structures through the prescribed placement of inorganic biological nanocomponents. However, challenge selectively growing DNA superlattices on specific locations surfaces their integration with conventional nanofabrication has hindered fabrication three-dimensional (3D) DNA-assembled functional devices. Here, we present a scalable technique that combines bottom-up top-down approaches...
Paul Roit, Johan Ferret, Lior Shani, Roee Aharoni, Geoffrey Cideron, Robert Dadashi, Matthieu Geist, Sertan Girgin, Leonard Hussenot, Orgad Keller, Nikola Momchev, Sabela Ramos Garea, Piotr Stanczyk, Nino Vieillard, Olivier Bachem, Gal Elidan, Avinatan Hassidim, Pietquin, Idan Szpektor. Proceedings of the 61st Annual Meeting Association for Computational Linguistics (Volume 1: Long Papers). 2023.
Josephson junctions are typically characterized by a single phase difference across two superconductors. This conventional two-terminal junction can be generalized to multiterminal device where the energy contains terms with contributions from multiple independent variables. Such (MTJJs) being considered as platforms for engineering effective Hamiltonians nontrivial topologies, such Weyl crossings and higher-order Chern numbers. These prospects rely on ability create MTJJs nonclassical...
The quest for the by-design assembly of material and devices from nanoscale inorganic components is well recognized. Conventional self-assembly often limited in its ability to control morphology structure simultaneously. Here, we report a general method assembling nanoparticles linear "pillar" with regulated internal configurations. Our approach inspired by supramolecular systems, where intermolecular stacking guides process form diverse morphologies. Programmable interactions were realized...
Mirror descent (MD), a well-known first-order method in constrained convex optimization, has recently been shown as an important tool to analyze trust-region algorithms reinforcement learning (RL). However, there remains considerable gap between such theoretically analyzed and the ones used practice. Inspired by this, we propose efficient RL algorithm, called {\em mirror policy optimization} (MDPO). MDPO iteratively updates approximately} solving problem, whose objective function consists of...
In this work, we describe a low-cost, two-step synthesis of composites nitrogen-doped carbon quantum dots (NCDs) with γ-Fe2O3 (NCDs/γ-Fe2O3), which is based on hydrothermal cum co-precipitation method. The product fine powder particles having an average diameter 9 ± 3 nm. physical and chemical properties NCDs/γ-Fe2O3 were studied, as well the superconducting interference device Mossbauer analysis magnetic these nanocomposites. interaction nanocomposites neuron-like cells was examined,...
Utilizing self-assembled DNA structures in the development of nanoelectronic circuits requires transforming strands into highly conducting wires. Toward this end, we investigate use nanowires as templates for deposition a superconducting material. Nanowires formed by NbN exhibit thermally activated and quantum phase slips well exceptionally large negative magnetoresistance. The latter effect can be utilized to suppress significant part low temperature resistance caused slips.
Abstract Quantum devices based on InSb nanowires (NWs) are a prime candidate system for realizing and exploring topologically-protected quantum states electrically-controlled spin-based qubits. The influence of disorder achieving reliable transport regimes has been studied theoretically, highlighting the importance optimizing both growth nanofabrication. In this work, we consider aspects. We developed NW with thin diameters, as well novel gating approach, involving few-layer graphene atomic...
Abstract Semiconductor nanowire (NW) quantum devices offer a promising path for the pursuit and investigation of topologically-protected states, superconducting spin-based qubits that can be controlled using electric fields. Theoretical investigations into impact disorder on attainment dependable topological states in semiconducting nanowires with large spin–orbit coupling g -factor highlight critical need improvements both growth processes nanofabrication techniques. In this work, we used...
Magnetic nanofluids are dispersions of magnetic nanoparticles in a diamagnetic base liquid, which display distinct physical properties that can be tuned easily by an external field, electric current, and temperature. Iron were synthesized sonochemically one-step process observed to oxidize situ over prolonged air exposure, forming α-Fe2O3 nanofluids. The thermal conductivity measurements on these single-step fabricated performed for the first time showed enhanced transport. Hence, we present...
Policy optimization methods are one of the most widely used classes Reinforcement Learning (RL) algorithms. Yet, so far, such have been mostly analyzed from an perspective, without addressing problem exploration, or by making strong assumptions on interaction with environment. In this paper we consider model-based RL in tabular finite-horizon MDP setting unknown transitions and bandit feedback. For setting, propose optimistic trust region policy (TRPO) algorithm for which establish $\tilde...
The Exploration-Exploitation tradeoff arises in Reinforcement Learning when one cannot tell if a policy is optimal. Then, there constant need to explore new actions instead of exploiting past experience. In practice, it common resolve the by using fixed exploration mechanism, such as $\epsilon$-greedy or adding Gaussian noise, while still trying learn an optimal policy. this work, we take different approach and study exploration-conscious criteria, that result policies with respect...
In Apprenticeship Learning (AL), we are given a Markov Decision Process (MDP) without access to the cost function. Instead, observe trajectories sampled by an expert that acts according some policy. The goal is find policy matches expert's performance on predefined set of functions. We introduce online variant AL (Online Learning; OAL), where agent expected perform comparably while interacting with environment. show OAL problem can be effectively solved combining two mirror descent based...
Reinforcement Learning from Human Feedback (RLHF) has become the standard approach for aligning Large Language Models (LLMs) with human preferences, allowing LLMs to demonstrate remarkable abilities in various tasks. Existing methods work by emulating preferences at single decision (turn) level, limiting their capabilities settings that require planning or multi-turn interactions achieve a long-term goal. In this paper, we address issue developing novel (RL) preference feedback between two...