A5029550449- Recommender Systems and Techniques
- Explainable Artificial Intelligence (XAI)
- Topic Modeling
- Complex Network Analysis Techniques
- Human Mobility and Location-Based Analysis
- Adaptive optics and wavefront sensing
- Fish Ecology and Management Studies
- Data Visualization and Analytics
- Natural Language Processing Techniques
- Ophthalmology and Visual Impairment Studies
- Data Management and Algorithms
- Machine Learning and Data Classification
- Data Quality and Management
- Scientific Computing and Data Management
- Advanced Bandit Algorithms Research
- Speech and dialogue systems
- Opportunistic and Delay-Tolerant Networks
- Peer-to-Peer Network Technologies
- Caching and Content Delivery
- Data Stream Mining Techniques
- Digital Marketing and Social Media
- Mobile Ad Hoc Networks
- Optical Wireless Communication Technologies
- Expert finding and Q&A systems
- Anomaly Detection Techniques and Applications
University Hospital Galway
2024
IBM Research - Ireland
2015-2023
Environment Canterbury
2023
Cambridge Scientific (United States)
2011-2021
Oregon State University
2019-2021
Oregon Department of Fish and Wildlife
2019-2021
IBM (United States)
2010-2020
The University of Texas at Austin
2019
Delft University of Technology
2018
Technology Centre Prague
2014
Message delivery in sparse Mobile Ad hoc Networks (MANETs) is difficult due to the fact that network graph rarely (if ever) connected. A key challenge find a route can provide good performance and low end-to-end delay disconnected where nodes may move freely. This paper presents multidisciplinary solution based on consideration of so-called small world dynamics which have been proposed for economy social studies recently revealed be successful approach exploited characterising information...
Message delivery in sparse mobile ad hoc networks (MANETs) is difficult due to the fact that network graph rarely (if ever) connected. A key challenge find a route can provide good performance and low end-to-end delay disconnected where nodes may move freely. We cast this as an information flow problem social network. This paper presents analysis metrics be used support novel practical forwarding solution efficient message delay-tolerant MANETs. These are based on of node's past interactions...
Discussion forums are a central part of Web 2.0 and Enterprise infrastructures. The health sustainability is dependent on the information exchange behaviour its contributors, which expressed through online conversation. increasing popularity importance requires better understanding characterisation communication so that can be managed, new services delivered opportunities risks detected. In this paper, we present an empirical analysis user roles in medium-sized bulletin board analyse...
The advent of real-time traffic streaming offers users the opportunity to visualise current conditions and congestion information. However, information highlighting underlying reason for tail-backs remains largely unexplored. Broken lights, an accident, a large concert, or road-works reveal important citizens operators alike. Providing such in requires intelligent mechanisms user interfaces order (i) harness heterogeneous data sources (volume, velocity, variety, veracity) (ii) make derived...
Chatbots or conversational recommenders have gained increasing popularity as a new paradigm for Recommender Systems (RS). Prior work on RS showed that providing explanations can improve transparency and trust, which are critical the adoption of RS. Their interactive engaging nature makes natural platform to not only provide recommendations but also justify through explanations. The recent surge interest inexplainable AI enables diverse styles justification, invites questions how...
Machine learning technologies are increasingly being applied in many different domains the real world. As autonomous machines and black-box algorithms begin making decisions previously entrusted to humans, great academic public interest has been spurred provide explanations that allow users understand decision-making process of machine model. Besides explanations, Interactive Learning (IML) seeks leverage user feedback iterate on an ML solution correct errors align with those users. Despite...
We present EvalAssist, a framework that simplifies the LLM- as-a-judge workflow. The system provides an online criteria development environment, where users can interactively build, test, and share custom evaluation in structured portable format. A library of LLM based evaluators is made available incorporates various algorithmic innovations such as token-probability judgement, positional bias checking, certainty estimation help to engender trust process. have computed extensive benchmarks...
Bringing a new AI system into production environment involves multiple different stakeholders such as business owners, risk officer, ethics officers approving the System for specific usage. Governance frameworks typically include manual steps, including curating information needed to assess risks and reviewing outcomes identify appropriate actions governance strategies. We demo human-in-the-loop automation that takes natural language description of an intended use case in order create...
The success of social media has resulted in an information overload problem, where users are faced with hundreds new contributions, edits and communications at every visit. A prime example this networks is the news or activity feeds, actions (friending, commenting, photo sharing, etc) friends on network presented to order inform them activity. In work we endeavour reduce burden individuals identifying interesting updates feeds by automatically recommending relevant items item relevance based...
Social networking sites have begun to be used in the enterprise as a method of connecting employees. Recommender systems may recommend social contacts order increase user engagement, encourage collaboration and facilitate expertise discovery. This paper evaluates effects four recommendation algorithms on network whole structure. We demonstrate that depending basis algorithm vary greatly their potential impact should understood. It is hoped this research can guidance for future algorithms.
The pyramid wavefront sensor is very similar to the Fourier knife-edge test, but employs dynamic modulation quantify phase derivative. For circular modulation, we compare approximate geometrical optics calculations, more exact diffraction and experimental results. We show that both sinusoidal linear relationship between derivative response can be derived rigorously from theory. also geometrical, results are similar, conclude predictions used in place of complex
The ever blurring line between online interactions and physical encounters presents an interesting challenge when recommending events. Events created on social networking sites may have ambiguous location scope. information provided be fuzzy or non existent additionally the reach radius of interest in event can vary greatly. In this work, we identify four categories events: global, dependent socially independent, dependent. We classify events from organizations internal management service...
The main applications of adaptive optics are the correction effects atmospheric turbulence on ground-based telescopes and ocular aberrations in retinal imaging visual simulation. requirements for wavefront corrector, usually a deformable mirror, will depend statistics to be corrected; here we compare spatial expected these two applications. We also use measured influence functions numerical simulations performance eight commercially available mirrors tasks. is studied as function size...
Large language models (LLMs) are susceptible to a variety of risks, from non-faithful output biased and toxic generations. Due several limiting factors surrounding LLMs (training cost, API access, data availability, etc.), it may not always be feasible impose direct safety constraints on deployed model. Therefore, an efficient reliable alternative is required. To this end, we present our ongoing efforts create deploy library detectors: compact easy-to-build classification that provide labels...
The photoluminescence (PL) of an ${\mathrm{In}}_{0.48}$${\mathrm{Ga}}_{0.52}$P/(${\mathrm{Al}}_{0.2}$${\mathrm{Ga}}_{0.8}$${)}_{0.52}$${\mathrm{In}}_{0.48}$P multiple-quantum-well sample composed wells various widths has been measured as a function temperature. presence LO-phonon replicas at low temperature for the largest well indicates that PL is dominated by localized excitons. This further confirmed variation peak energies and linewidths increased above 4.2 K. dependence integrated...
Efficiently detecting conversation threads from a pool of messages, such as social network chats, emails, comments to posts, news etc., is relevant for various applications, including Web Marketing, Information Retrieval and Digital Forensics. Existing approaches focus on text similarity using keywords features that are strongly dependent the dataset. Therefore, dealing with new corpora requires further costly analyses conducted by experts find out features. This paper introduces...
We describe the operation of a pyramid wavefront sensor used to measure and correct aberrations human eye. The system is designed for maximum speed when running in closed loop but can also provide calibrated open-loop measurements with range sampling options. A detailed characterization was performed ensure measurement accuracy. Ocular after correction had root-mean-square errors consistently less than 0.1μm over 6mm pupil all subjects tested. frame rate 83Hz both open- closed-loop modes.
We present AutoDOViz, an interactive user interface for automated decision optimization (AutoDO) using reinforcement learning (RL). Decision (DO) has classically being practiced by dedicated DO researchers where experts need to spend long periods of time fine tuning a solution through trial-and-error. AutoML pipeline search sought make it easier data scientist find the best machine leveraging automation and tune solution. More recently, these advances have been applied domain AutoDO, with...
This paper reports the findings of Dagstuhl Perspectives Workshop 17442 on performance modeling and prediction in domains Information Retrieval, Natural language Processing Recommender Systems. We present a framework for further research, which identifies five major problem areas: understanding measures, analysis, making underlying assumptions explicit, identifying application features determining performance, development models describing relationship between assumptions, resulting performance.