Category: Announcements

November 7, 202412:15 pm - 1:15 pm Speaker:Tony Mangino, Department of Biostatistics, University of Kentucky Where:327 McVey Hall(Zoom link: https://uky.zoom.us/j/82467171189) Title:Using the RGAN Package for Implementing Generative Models for Biomedical Applications Abstract: While generative adversarial networks (GANs) are widely implemented using a variety of software packages and programming languages, many applied researchers make use of the R Statistical Software Package to conduct their analyses. The development of the RGAN package (Neunhoeffer, 2022) now allows for a user-friendly command line interface for training GANs on tabular or image data in the R environment. This tutorial guides attendees through the process of training and evaluating a GAN trained using the RGAN framework through a real clinical example dataset. Attendees who wish to follow along must have the following software installed: The R…

October 31, 202412:15 pm - 1:15 pm Speaker:Tony Mangino, Department of Biostatistics, University of Kentucky Where:327 McVey Hall(Zoom link: https://uky.zoom.us/j/82467171189) Title:An introduction to generative AI in Biomedical Applications Abstract: Generative AI has become ubiquitous since the advent of ChatGPT, though the underlying mechanisms are far from novel. Goodfellow’s 2014 introduction of generative adversarial networks (GANs) was a watershed moment for the generative modeling domain and has since been widely used in a wide variety of contexts. This presentation and accompanying tutorial introduce participants to a lesser-used application of GANs: Their ability to generate synthetic patient records with data largely mirroring the same data obtained from real patients. This presentation illustrates an example using data from patients admitted to UK hospitals for either Acute Coronary Syndrome (ACS) or Takotsubo Syndrome (TTS),…

October 24, 202412:15 pm - 1:15 pm Speaker:Qiang Ye, Department of Mathematics, University of Kentucky Where:327 McVey Hall(Zoom link: https://uky.zoom.us/j/82467171189) Title:Recurrent Neural Networks and Transformer for Sequential Data Abstract: Many machine learning problems involve sequential data. Recurrent neural networks (RNNs) and Transformer are neural network architectures designed to efficiently model temporal connections within a sequence and handling variable sequence lengths in a dataset. However, RNNs suffer from the so-called vanishing or exploding gradient problems, which also reduces its ability to pass information in a long sequence. Transformer solves this problem through a self-attention mechanism but faces challenges in efficiently scaling to long sequences because the self-attention computation is quadratic with respect to the sequence length. We will present several orthogonal RNN models that we have developed to address the vanishing/exploding…

October 17, 202412:15 pm - 1:15 pm Speaker:Aaron Mullen, Institute for Biomedical Informatics, University of Kentucky Where:VIRTUAL!(Zoom link: https://uky.zoom.us/j/82438134047) Title:Forecasting Opioid Incidents for Rapid Actionable Data for Opioid Response in Kentucky Abstract: We present efforts in the fields of machine learning and time series forecasting to accurately predict counts of future opioid overdose incidents around the state of Kentucky. If successful, state governments could use forecasts to properly prepare and distribute resources effectively. The approach taken primarily uses county and district level aggregations of EMS opioid overdose encounters and forecasts future counts at a monthly level. A variety of additional covariates were also tested to determine their impact on the model’s performance. Models with different levels of complexity were evaluated as well to optimize training time and accuracy. The results…

October 10, 202412:15 pm Speaker:Md Atik Ahamed, Division of Biomedical Informatics, University of Kentucky Where:327 McVey Hall(Zoom link: https://uky.zoom.us/j/82467171189) Title:Advanced tools and packages for handling tabular data and time series data Abstract: In this comprehensive tutorial, we will assess several state-of-the-art models for analyzing both tabular and time series data. We will begin by discussing the advantages and limitations of models such as XGBoost, TransTab and MambaTab for tabular data, as well as deep learning approaches like PatchTST and TimeMachine models for time series prediction.Following this overview, we will provide a step-by-step walkthrough of how to use corresponding tools and packages effectively. We will focus on real-world datasets and explanations of the packages in a user-friendly manner. Click here to see the complete list of speakers.

October 3, 202412:15 pm Speaker:Qiang (Shawn) Cheng, Division of Biomedical Informatics, University of Kentucky Where:327 McVey Hall(Zoom link: https://uky.zoom.us/j/82467171189) Title:Novel Deep Learning models and Techniques for Effective and EfficientHandling of Multiple Data Modalities Abstract: Diverse areas of scientific research and everyday life, including healthcare, biomedicine, and engineering, are inundated with various data modalities, each presenting unique challenges. This talk presents several cutting-edge learning approaches designed to handle different types of data with both accuracy and computational efficiency. They include A SOTA state-space model for tabular data classification, time series long-term forecasting with linear complexity, time series classification, multi-modal data clustering with sparse tensor learning, and if time permits inferring circadian phases from multi-omics data. This talk integrates deep learning models and classical learning methods, each optimized for a particular data…

September 26, 202412:15 pm Speaker:Aaron Mullen, Institute for Biomedical Informatics, University of Kentucky Where:327 McVey Hall(Zoom link: https://uky.zoom.us/j/82467171189) Title:CLASSify: A Self-Service Machine Learning Platform Abstract: CLASSify is a web-based tool developed at the Center for Applied AI to make machine learning easier and more accessible. It provides a platform to train and evaluate ML classification models on any tabular data without requiring any programming background. Users can simply upload their dataset to the site, choose the training parameters, and the job will be sent off to train all chosen models and provide results in the form of tables and visualizations. CLASSify also provides options for synthetic data generation to bolster imbalanced class labels or create entirely new datasets, as well as explainability scores that provide insight into which features of…

September 19, 202412:15 pm Speaker:Brent Harrison, Computer Science, University of Kentucky Where:327 McVey Hall(Zoom link: https://uky.zoom.us/j/82467171189) Title:Practical Value Alignment Using GPT-4o Abstract: As artificial intelligence (AI) and machine learning (ML) systems grow in power, their risk of causing unintentional harm grows. This is because often these systems are optimizing for criteria that is different than our own, often meaning that it does not consider the many social and cultural norms that we implicitly and explicitly use when making decisions. In a value-aligned system, however, system behavior and objectives are aligned with expected human sociocultural norms. Creating a value-aligned system, however, can be difficult as sources of explicit value information are scarce. In this tutorial, I will discuss how large language models, specifically GPT-4o, can be used to perform practical value…

September 12, 202412:15 pm Speaker:Brent Harrison, Computer Science, University of Kentucky Where:327 McVey Hall(Zoom link: https://uky.zoom.us/j/82467171189) Title:A Gentle Introduction to Modern Machine Learning Abstract: Machine learning has rapidly grown over the last few years. Currently, there are many elements of our social and work lives that utilize, in some way, a machine learning system. To better utilize these systems, it is important to understand how these systems work. In this talk, I will provide an overview of machine learning systems to provide a foundation that the other talks in this series can build on. I will focus on high-level details of how these systems work, how they're often used in practice, and their data requirements. Click here to see the complete list of speakers.

May 7, 20243:00 pm – Refreshments3:30 pm – Presentation Speaker:Emilio Romano-Diaz, Argelander-Institute for Astronomy, University of Bonn, Germany Where:Davis Marksbury Building – James F. Hardymon Theatre(Zoom link: https://uky.zoom.us/s/84474671604) Title:Cosmology via big data in Astrophysics  Abstract: State-of-the-art observational surveys such as the one carried out by the EUCLID mission, BOSS, DES and forthcoming ones like LSST, HETDEX among others, will cover of the order of 10000 square degrees on the sky, with the primary science goal to unravel the nature of the physics responsible for the current accelerated expansion of the universe. The unprecedented and rich data provided by these surveys will make it possible to investigate fundamental physics (e.g. inflation, neutrino properties) and astrophysics (e.g. biasing, galaxy formation). The success of future large-scale galaxy surveys evidently requires a correct interpretation…