A5100629989- Low-power high-performance VLSI design
- Caching and Content Delivery
- Gaze Tracking and Assistive Technology
- Seismic Waves and Analysis
- Age of Information Optimization
- Subtitles and Audiovisual Media
- Adversarial Robustness in Machine Learning
- Advanced Causal Inference Techniques
- Digital Mental Health Interventions
- Optical measurement and interference techniques
- Anomaly Detection Techniques and Applications
- IoT and Edge/Fog Computing
- Privacy, Security, and Data Protection
- IoT Networks and Protocols
- Digital Media Forensic Detection
- Visual Attention and Saliency Detection
- Peer-to-Peer Network Technologies
- Information Retrieval and Data Mining
- Network Traffic and Congestion Control
- Generative Adversarial Networks and Image Synthesis
- Statistical Methods in Clinical Trials
- Video Surveillance and Tracking Methods
- Sexuality, Behavior, and Technology
- Image and Video Quality Assessment
- VLSI and Analog Circuit Testing
University of Illinois Urbana-Champaign
2023-2025
Technology Information, Forecasting and Assessment Council
2022
University of Wisconsin–Madison
2020
Lakehead University
2020
University of North Carolina at Greensboro
2011
San Francisco State University
2011
Extended reality (XR) devices, including augmented, virtual, and mixed reality, provide a deeply immersive experience. However, practical limitations like weight, heat, comfort put extreme constraints on the performance, power consumption, image quality of such systems. In this paper, we study how these form tradeoff between Fixed Foveated Rendering (FFR), Gaze-Tracked (TFR), conventional, non-foveated rendering. While existing papers have often studied methods, first comprehensive their...
We design and implement a personalized automated physical activity coaching engine, PACE, which uses the Fogg's behavioral model (FBM) to engage users in mini-conversation based sessions. It is chat-based nudge assistant that can boost (encourage) sense (ask) motivation, ability propensity of walk help them achieving their step count targets, similar human coach. demonstrate feasibility, effectiveness acceptability PACE by directly comparing coaches Wizard-of-Oz deployment study with 33...
Many causal and structural parameters are linear functionals of an underlying regression. The Riesz representer is a key component in the asymptotic variance semiparametrically estimated functional. We propose adversarial framework to estimate using general function spaces. prove nonasymptotic mean square rate terms abstract quantity called critical radius, then specialize it for neural networks, random forests, reproducing kernel Hilbert spaces as leading cases. Furthermore, we use radius...
We address the problem of optimal routing in overlay networks. An network is constructed by adding new nodes on top a legacy network. The are capable implementing any dynamic policy, however, underlay has fixed, single path scheme and uses simple work-conserving forwarding policy. Moreover, routes pre-determined unknown to can increase achievable throughput using multiple routes, which consist direct indirect through other nodes. develop algorithm for such networks called Optimal Overlay...
We consider the problem of transmission scheduling for remote estimation a discrete-time autoregressive Markov process that is driven by white Gaussian noise. A sensor observes this process, and then decides to either encode current state into data packet attempts transmit it estimator over an unreliable wireless channel modeled as Gilbert-Elliott channel, or does not send any update. Each attempt consumes $\lambda$ units power, assumed be linear. The revealed only via feedback (ACK\slash...
The objective of this research project is to create a tool that can transform written material into an audio format, thereby enhancing accessibility for people who are visually impaired or illiterate.The will employ Optical Character Recognition (OCR) recognize and extract text from images, which then be converted document.Subsequently, the document undergo Text-to-Speech (TTS) conversion, resulting in file.To achieve goal, system execute series image processing operations text, transformed...
Applications suited to approximation often exhibit significant value locality, both in terms of inputs as well outcomes. In this early stage proposal - the ODIN: Outcome Driven Input Navigated approach locality based approximation, we hypothesize that optimizations for approximate applications should be driven by outcomes i.e., result computation, but navigated with help inputs. An outcome-driven can enable computation slices, whose are deemed (approximately) redundant or derivable, entirely...
<p>Civic technology is a fast-developing segment that holds huge potential for new generation of startups. A recent survey report on civic noted the sector saw $430 million in investment just last two years. It's not market ripe with opportunity it's crucial to our democracy. Crowdsourcing has proven be an effective supplementary mechanism public engagement city government order use mutual knowledge online communities address such issues as means engaging people urban design....
We consider the problem of content caching at wireless edge to serve a set end users via unreliable channels so as minimize average latency experienced by due constrained cache capacity. formulate this Markov decision process, or more specifically restless multi-armed bandit problem, which is provably hard solve. begin investigating discounted counterpart, and prove that it admits an optimal policy threshold-type. then show result also holds for problem. Using structural result, we establish...
We extend the idea of automated debiased machine learning to dynamic treatment regime and more generally nested functionals. show that multiply robust formula for with discrete treatments can be re-stated in terms a recursive Riesz representer characterization mean regressions. then apply estimation algorithm estimates de-biasing corrections without need characterize how correction look like, such as instance, products inverse probability weighting terms, is done prior work on doubly regime....
Because of the availability larger datasets and recent improvements in generative model, more realistic Deepfake videos are being produced each day. People consume around one billion hours video on social media platforms every day, thats why it is very important to stop spread fake as they can be damaging, dangerous, malicious. There has been a significant improvement field deepfake classification, but detection inference have remained difficult task. To solve this problem paper, we propose...