events

The Proximal Distance Principle for Constrained Estimation

tmaju002 — Tue, 28 May 2024 14:46:13 +0000

The Proximal Distance Principle for Constrained Estimation tmaju002 Tue, 05/28/2024 - 07:46 More News May 24, 2024

Statistical methods often involve solving an optimization problem, such as in maximum likelihood estimation and regression. The addition of constraints, either to enforce a hard requirement in estimation or to regularize solutions, complicates matters. Fortunately, the rich theory of convex optimization provides ample tools for devising novel methods. In this talk, I present applications of distance-to-set penalties to statistical learning problems. Specifically, I will focus on proximal distance algorithms, based on the MM principle, tailored to various applications such as regression and discriminant analysis. Special emphasis is given to sparsity set constraints as a compromise between exhaustive combinatorial searches and lasso penalization methods that induce shrinkage.

Dr. Alfonso Landeros

Tags

events

Data-Efficient Deep Learning using Physics-Informed Neural Networks

tmaju002 — Tue, 28 May 2024 14:39:37 +0000

Data-Efficient Deep Learning using Physics-Informed Neural Networks tmaju002 Tue, 05/28/2024 - 07:39 More News May 10, 2024

A grand challenge with great opportunities is to develop a coherent framework that enables blending conservation laws, physical principles, and/or phenomenological behaviours expressed by differential equations with the vast data sets available in many fields of engineering, science, and technology. At the intersection of probabilistic machine learning, deep learning, and scientific computations, this work is pursuing the overall vision to establish promising new directions for harnessing the long-standing developments of classical methods in applied mathematics and mathematical physics to design learning machines with the ability to operate in complex domains without requiring large quantities of data. To materialize this vision, this work is exploring two complementary directions: (1) designing data-efficient learning machines capable of leveraging the underlying laws of physics, expressed by time dependent and non-linear differential equations, to extract patterns from high-dimensional data generated from experiments, and (2) designing novel numerical algorithms that can seamlessly blend equations and noisy multi-fidelity data, infer latent quantities of interest (e.g., the solution to a differential equation), and naturally quantify uncertainty in computations.

Dr. Maziar Raissi

Tags

events

Electronic Bee-Veterinarian: A Data Centric Approach to Monitor Honeybee Health

tmaju002 — Tue, 28 May 2024 14:34:51 +0000

Electronic Bee-Veterinarian: A Data Centric Approach to Monitor Honeybee Health tmaju002 Tue, 05/28/2024 - 07:34 More News April 26, 2024

Honeybees are vital for pollination and food production. Among many factors, extreme temperature (e.g., due to climate change) is particularly dangerous for bee health. Anticipating such extremities would allow beekeepers to take early preventive action. Thus, given sensor (temperature) time series data from beehives, how can we find patterns and do forecasting? Forecasting is crucial as it helps spot unexpected behavior and thus issue warnings to the beekeepers. In that case, what are the right models for forecasting? ARIMA, RNNs, or something else?

We propose the EBV (Electronic Bee-Veterinarian) method, which has the following desirable properties: (i) principled: it is based on a) diffusion equations from physics and b) control theory for feedback-loop controllers; (ii) effective: it works well on multiple, real-world time sequences, (iii) explainable: it needs only a handful of parameters (e.g., bee strength) that beekeepers can easily understand and trust, and (iv) scalable: it performs linearly in time. We applied our method to multiple real-world time sequences, and found that it yields accurate forecasting (up to 49% improvement in RMSE compared to baselines), and segmentation. Specifically, discontinuities detected by EBV mostly coincide with domain expert's opinions, showcasing our approach's potential and practical feasibility. Moreover, EBV is scalable and fast, taking about 20 minutes on a stock laptop for reconstructing two months of sensor data. (See paper at: https://epubs.siam.org/doi/epdf/10.1137/1.9781611978032.34)

Joint work with Christos Faloutsos (CMU), Boris Baer (UCR), Hyoseung Kim (UCR), and Vassilis Tsotras (UCR)

Mst. Shamima Hossain

Tags

events

LAMBRETTA: Learning to Rank for Twitter Soft Moderation

tmaju002 — Tue, 28 May 2024 14:19:23 +0000

LAMBRETTA: Learning to Rank for Twitter Soft Moderation tmaju002 Tue, 05/28/2024 - 07:19 More News April 19, 2024

To curb the problem of false information, social media platforms like Twitter started adding warning labels to content discussing debunked narratives, with the goal of providing more context to their audiences. Unfortunately, these labels are not applied uniformly and leave large amounts of false content unmoderated. This talk presents LAMBRETTA, a system that automatically identifies tweets that are candidates for soft moderation using Learning To Rank (LTR). We run LAMBRETTA on Twitter data to moderate false claims related to the 2020 US Election and find that it flags over 20 times more tweets than Twitter, with only 3.93% false positives and 18.81% false negatives, outperforming alternative state-of-the-art methods based on keyword extraction and semantic search.

Overall, LAMBRETTA assists human moderators in identifying and flagging false information on social media (PDF: IEEE Xplore Full-Text PDF:)

Dr. Emiliano De Cristofaro

Tags

events

Charting the Future: Integrating Health Informatics into Undergraduate Medical Education

tmaju002 — Tue, 28 May 2024 14:13:55 +0000

Charting the Future: Integrating Health Informatics into Undergraduate Medical Education tmaju002 Tue, 05/28/2024 - 07:13 More News April 12, 2024

Digital information systems have become central to the practice of medicine over the last twenty years, and the next twenty years will bring even greater changes challenges, and benefits for physicians, patients, and health systems. While trends in AI, Big Data, and Digital Health indicate major changes on the horizon, health informatics platforms that derive data from clinical environments have already begun to change how physicians and public health research real-world clinical problems. In recognition of the value of these technologies, the American Medical Association has ascribed Health informatics a prominent place as a new competency in the field of Health Systems Science. As students begin their journey towards becoming physician leaders, they will need the knowledge, skills, and critical perspectives necessary to understand, use, and steward these technological systems even as they continue to evolve.

In this talk, I will describe the current need for health informatics education in undergraduate medical education (MD) programs, how these technologies can dovetail with the pressing need to standardize and improve research education in UME, and provide a demo of the TriNetX clinical data platform that will be available at UCR starting on April 15, 2024. Participant questions will be welcomed.

Dr. Daniel Novak

Tags

events

Using AI to enhance FIB-SEM Image Segmentation For Cell Biology

tmaju002 — Tue, 28 May 2024 14:09:20 +0000

Using AI to enhance FIB-SEM Image Segmentation For Cell Biology tmaju002 Tue, 05/28/2024 - 07:09 More News March 08, 2024

Focused Ion Beam Scanning Electron Microscopy (FIB-SEM) has been improved to the point that nanoscale imaging of whole cell and even larger biological samples can now be reliably obtained. The science bottleneck now moves to processing these images to extract insights and knowledge. Librarians at UCR and Virginia Tech are working with microscopy and biomed researchers at Yale School of Medicine to segment cell organelles from recently acquired enhanced FIB-SEM images of mouse neurons. While we did not develop new machine-learning algorithms, practical AI work still requires smart strategies to apply these algorithms to their fullest effects. We will discuss these practical considerations, from ground truth labeling, training, and post-processing, and highlight the necessity to adapt machine-learning pipeline to the needs of the scientific drives.

This project showcases how librarians can be embedded and contribute to faculty research. UCR Library's Research Data Initiative is open to collaborate with UCR faculty on data intensive and AI assisted methods to directly help solve domain research problems. Come and learn about the UCR library's vision on this initiative, and solicit potential collaborative project ideas.

Dr. Zhiwu Xie

Tags

events

Dealing with Latent Pre-exposure to Information Treatments

tmaju002 — Sun, 26 Feb 2023 08:38:51 +0000

Dealing with Latent Pre-exposure to Information Treatments tmaju002 Sun, 02/26/2023 - 00:38 More News March 03, 2023

In Social Sciences, many experiments rely on responses to information treatments. Experimental subjects in the treatment group receive some information that subjects in the control group don't. Often, the proportion of people in the treatment and control groups who were pre-exposed to the information is unknown and uncontrolled by the researchers. If that pre-exposure is ignored, it can bias the treatment effect estimation and lead to incorrect conclusions. I propose two estimation procedures for latent pre-exposure in this paper. One combines designed-based and model-based methods and relies on ancillary sampling and predictive probability of pre-exposure. The other is data-driven and relies on exploring latent effect heterogeneity using unsupervised learning methods and Dirichlet Process models. I compare both approaches using a real application to investigate public attitudes toward government tax policies to reduce economic inequality.

Dr. Diogo Ferrari

Tags

events

Machine Learning Guided Modeling of Ligand-Protein Binding Energy Landscape: Applications in Small Molecule and Protein-based Drug Design.

tmaju002 — Sun, 26 Feb 2023 08:35:20 +0000

Machine Learning Guided Modeling of Ligand-Protein Binding Energy Landscape: Applications in Small Molecule and Protein-based Drug Design. tmaju002 Sun, 02/26/2023 - 00:35 More News February 17, 2023

Molecules in cells constantly move. The motions of proteins in living cells can be simple fluctuations or functional. Therefore, investigating protein dynamics is crucial for understanding protein function and for accurately compute ligand-protein binding free energy landscape. Because experimental structures are static conformations, classical or enhanced molecular dynamics (MD) simulations are commonly used for conformational sampling. Machine/deep learning approaches can then be used to analyze MD results and assist conformational sampling and energy calculations.

In this presentation, we will focus on modeling ligand-receptor binding/unbinding pathways to compute protein-drug binding thermodynamics and kinetics for drug development. We will show the binding free energy landscape constructed by Binding Kinetics Toolkit (BKiT), a program using post-analysis, principal component analysis and milestoning theory to predict drug binding kinetics. We will also discuss use of machine learning and deep learning to enhance protein conformational sampling to model protein conformational transition and other applications.

Dr. Chia-en Chang

Tags

events

Some Thoughts on Data Science - Population Health Collaborations.

tmaju002 — Sun, 26 Feb 2023 08:28:51 +0000

Some Thoughts on Data Science - Population Health Collaborations. tmaju002 Sun, 02/26/2023 - 00:28 More News February 10, 2023

Data science involves the application of knowledge from the fields of computer science (on how to manage data) and statistics (on how to analyze data) to solve theoretical and practical problems. The field of population health involves investigation of health outcomes, patterns of health determinants, and policies and interventions that link them (Kindig and Stoddart, 2003). Data science and population health have a natural affinity. This talk will explore this affinity, including potential avenues for realizing its potential. Special reference will be made to opportunities at UCR for collaborations of data scientists and population health scholars involving research, teaching, and service.

Dr. Mark Wolfson

Tags

events

Remote Sensing of plant and soil for precision agriculture

tmaju002 — Thu, 02 Feb 2023 19:51:50 +0000

Remote Sensing of plant and soil for precision agriculture tmaju002 Thu, 02/02/2023 - 11:51 More News February 03, 2023

Agricultural systems are often characterized by high spatial and temporal variability in the factors that determine crop yield. In particular, the variability of soil and other environmental factors affecting yield are notoriously hard to characterize at very high spatial resolution. Recent high-resolution satellites (e.g., Sentinel and PlanetScope) may be useful tools for monitoring crops and environmental factors in agriculture. This talk will describe recent research in Scudiero’s Lab on remote sensing of soil moisture and on the determination of crop salinity tolerance in crops using high-resolution satellite data. The talk will include an overview of research gaps and opportunities for the use of data science in agricultural research.

Dr. Elia Scudiero

Tags

events