AI Innovation Institute

Virtual Talk: Metadata Matters: Robust Document Classification via Adaptation Methods for Text-driven Public Health by Xiaolei Huang

Zoom link to follow.

Abstract: Document classifiers have been widely applied in solving health-related issues, such as suicide prevention, flu vaccination surveillance and disease diagnosis. However, document metadata including time, gender, age and location has an enormous impact on robustness of 
document classifiers. Language varies across the metadata bringing both challenges and opportunities to build reliable document classifiers. For example, online written language changes over time, and males and females express opinions differently. This talk describes how to use domain adaptation to integrate temporal and user demographic factors into document classifiers. By adapting knowledge of how language varies across the metadata, models can learn generalized representations of language through the metadata-invariant embeddings. 
This approach will lead to metadata-adapted document classifiers and can also extend to personalize classification models by user embedding. 

Bio: Xiaolei Huang is a 4th-year PhD candidate in Information Science at the University of Colorado, Boulder. He is currently a visiting scholar at the Johns Hopkins University. His research interests are in Natural Language Processing, Machine Learning and Public Health. Particularly, he focuses on domain adaptation, cross-lingual transfer learning, user modeling and fairness.

Abstract: Gaussian Probability Path-based Generative Models (GPPGMs) generate data by reversing a stochastic process that progressively corrupts samples with Gaussian noise. Despite state-of-the-art results in 3D molecular generation, their deployment is hindered by the high cost of long generative trajectories, often requiring hundreds to thousands of steps during training and sampling. In this work, we propose a principled method, named GAGA, to improve generation efficiency without sacrificing training granularity or inference fidelity of GPPGMs. Our key insight is that different data modalities obtain sufficient Gaussianity at markedly different steps during the forward process. Based on this observation, we analytically identify a characteristic step at which molecular data attains sufficient Gaussianity, after which the trajectory can be replaced by a closed-form Gaussian approximation. Unlike existing accelerators that coarsen or reformulate trajectories, our approach preserves full-resolution learning dynamics while avoiding redundant transport through truncated distributional states. Experiments on 3D molecular generation benchmarks demonstrate that our GAGA achieves substantial improvement on both generation quality and computational efficiency.

Speaker: Jingxiang Qu

Location: New Computer Science 220

Abstract: Modern language agents often need to solve tasks requiring long-horizon, multi-turn interactions, where they retrieve external information, adapt to observations, and answer interdependent queries. Yet, most LLM systems rely on full-context prompting, appending all past turns regardless of their relevance. This leads to un-bounded memory growth, increased computational costs, and degraded reasoning performance on out-of-distribution input lengths due to LLM forgetting the context. We introduce MEM1, an end-to-end reinforcement learning framework that enables agents to operate with constant context size when solving long multi-turn tasks. At each turn, MEM1 updates a compact shared internal state that jointly supports memory consolidation and reasoning. Leveraging reinforcement learning (RL) and rollout trajectory truncation, we train a MEM1 agent to develop internal states that integrate prior memory with new observations from the environment while strategically discarding irrelevant or redundant information. Experiments across three domains, including internal retrieval QA, open-domain web QA, and multi-turn web shopping, show that MEM1-7B improves performance by 3.5x while reducing memory usage by 3.7x compared to Qwen2.5-14B-Instruct on an augmented multi-hop QA dataset with 16 objectives in each task, and generalizes beyond the training horizon. Our results demonstrate the promise of reasoning-driven memory consolidation as a scalable alternative to existing solutions for training long-horizon task-solving agents that involve multiple interactions, where both efficiency and performance are optimized.

Speaker: Yiyang Feng

Location: CS2311

Abstract: In today's digital era, language functions not only as a medium of information transmission but also as a mechanism of persuasion, framing, and control. The proliferation of online platforms has amplified this dual role: while enabling unprecedented access to knowledge, it has also exacerbated challenges such as misinformation, rhetorical manipulation, and cultural or linguistic disparities in information access. As a result, pragmatic language understanding and information integrity have emerged as central concerns for both computational linguistics and society at large. This research follows how claims are produced, reframed, and contested online through three interconnected threads. First, it models pragmatic deflection in discourse by investigating whataboutism, a rhetorical device that deflects criticism by redirecting discourse, and introduced novel datasets from Twitter (now X) and YouTube. This work underscores how subtle pragmatic maneuvers can erode discourse integrity without relying on outright falsehoods. Second, it advances retrieval and alignment for information integrity in health and news communication. These systems trace claims and narratives across genres (e.g., social posts and news reports) and languages (Chinese and English), linking social posts with journalistic reporting and aligning Chinese news with English biomedical evidence. By accounting for cultural context, assertions can be linked to reliable evidence and organized for systematic comparison. This work surfaces the risks of missing sources, unverifiable claims, and framing disparities in global health discourse, and demonstrates computational solutions that enhance both the credibility and accessibility of information. Third, the methodological centerpiece is Class Distillation (ClaD), a geometry-aware training paradigm for distilling a small, well-defined target class from a large, heterogeneous background. ClaD couples a distribution-aware contrastive loss (instantiated here in a Mahalanobis form when its assumptions fit the data) with an interpretable decision algorithm tuned for class separation. Evaluated on sarcasm, metaphor, and sexism detection, ClaD delivers strong efficiency and robustness, matching or surpassing larger models while using fewer computational resources, making these pipelines practical by learning reliably from small, sharply defined classes. In sum, this research presents an integrated account of language understanding in the digital age. It exposes how integrity falters through pragmatic deflection, cross-genre drift, and cross-lingual misalignment, and translates these insights to move pragmatic language understanding to systems for evidence retrieval, alignment, and verification; and it sheds light on where and how integrity is threatened, and delivers methods that leverage pragmatic language use.

Speaker: Chenlu Wang

Location: (Old) Computer Science Building, Room 2311

Abstract: The recent expansion of online sport wagering and igaming has led to higher rates of problem gambling, particularly among emerging adults and other population subgroups. The Center for Gambling Studies (CGS) at the Rutgers University, School of Social Work, is using big data analysis, machine learning and GIS mapping to identify geographic locations with populations most at risk to guide the development of targeted interventions. This presentation will review the GIS StoryMap for the State of New Jersey, including a blueprint for the highest risk target service areas in the state. It will also present findings from a machine learning model that identifies the key risk factors for high-intensity online casino bettors. Implications for prevention, treatment and policy initiatives will be discussed.

Bio: Lia Nower, J.D., Ph.D., is a Distinguished Professor, Associate Dean for Research, and Director of the Center for Gambling Studies at Rutgers University. A clinician and attorney, her research focuses on big data analysis and machine learning models for online gambling and sports wagering; gambling and video gaming among emerging adults; policy initiatives around harm reduction and responsible gambling, and etiology and treatment of problem gambling. Dr. Nower serves as a senior editor for Addiction. She has received both the Research (2019) and the Lifetime Research Award (2022) from the National Council on Problem Gambling and the Board of Trustees Award for Research (2022) from Rutgers University.

Join Zoom Meeting: https://stonybrook.zoom.us/j/95617197636?pwd=KytzZ2pVRG9SZGpKZUtpNXJISjNjZz09
Meeting ID: 956 1719 7636 Passcode: 924293

Abstract:

Capturing the spatio-temporal (4D) dynamics of humans has been a long standing research problem in computer vision and graphics. Synthesizing photorealistic human avatars has broad applications, ranging from immersive telepresence in AR/VR and the movie industry, to enriching the education and healthcare systems. Earlier approaches relied on hand-engineered models that use a small amount of data from one or more subjects. With the advent of neural networks, training on large datasets enhanced the output visual quality. Currently, the combination of neural networks with graphics techniques has achieved natural-looking human animation. However, most approaches are identity-specific, trained only on a single identity, and use only one modality.

In this thesis, we address the problem of learning neural representations of humans in a holistic way. Given that the video data in the real world include multiple modalities (audio and video) and multiple identities, we develop multi-modal and multi-identity representations. First, we propose to reconstruct the 4D face geometry of humans by leveraging both audio and video information. In this way, the network produces accurate lip shapes and is robust to cases when either modality is insufficient. Next, we introduce a NeRF-based representation for audio-driven human face animation that achieves high-quality lip synchronization for cinematic content. Since humans communicate with their full body, combining body pose, hand gestures, and facial expressions, we extend our network to capture the full-body human motion for multiple identities simultaneously. In order to better disentangle identity and non-identity specific information, we subsequently study non-linear interactions between latent factors of variation, and propose a specific multiplicative module. In this way, we learn a multi-identity NeRF that robustly animates human faces under novel expressions and achieves a significant decrease in the total training time. Similarly, we propose a multi-identity gaussian splatting representation for human bodies, by constructing a high-order tensor. Assuming a low-rank structure, we learn a tensor decomposition that leads to a significant decrease in the total number of learnable parameters, as well as to a robust animation under novel poses. In the future, we propose to jointly synthesize audio and visual outputs from just text input. Given the recent rise of large language models, coupling text with natural-looking avatars can enhance the overall interaction between a human and an AI system.

Speaker: Aggelina Chatziagapi

Where: NCS, Room 220

Zoom link: https://stonybrook.zoom.us/j/98775312249?pwd=uORNAnSdcssrPZdqOsqaMAF5aLcRD9.1
ID: 98775312249
Passcode: 505777

Nam Nguyen

4-5pm, Dec 17 2020

https://stonybrook.zoom.us/j/94214254415?pwd=K1VoQml4cFdlVW51VW41dWtid2tJdz09



The molecular mechanisms and functions in complex biological systems
currently remain elusive. Recent high-throughput techniques, such as
next-generation sequencing, have generated a wide variety of
multiomics datasets that enable the identification of biological
functions and mechanisms via multiple facets. However, integrating
these large-scale multiomics data and discovering functional insights
are, nevertheless, challenging tasks. To address these challenges,
machine learning has been broadly applied to analyze multiomics. In
particular, multiview learning is more effective than previous
integrative methods for learning data's heterogeneity and revealing
cross-talk patterns. Although it has been applied to various contexts,
such as computer vision and speech recognition, multiview learning has
not yet been widely applied to biological data--specifically,
multiomics data. Therefore, we have developed a framework called
multiview empirical risk minimization (MV-ERM) for unifying multiview
learning methods (Nguyen, et al., PLoS Computational Biology, 2020).
MV-ERM enables potential applications to understand multiomics
including genomics, transcriptomics, and epigenomics, in an aim to
discover the functional and mechanistic interpretations across omics.
Based on MV-ERM, we have developed the following methods:
ManiNetCluster, Varmole and ECMarker.



(1) ManiNetCluster (Nguyen, et al., BMC Genomics, 2019) is a manifold
learning method which simultaneously aligns and clusters gene networks
(e.g., co-expression) to systematically reveal the links of genomic
function between different phenotypes. Specifically, ManiNetCluster
employs manifold alignment to uncover and match local and non-linear
structures among networks, and identifies cross-network functional
links. We demonstrated that ManiNetCluster better aligns the
orthologous genes from their developmental expression profiles across
model organisms than state-of-the-art methods. This indicates the
potential non-linear interactions of evolutionarily conserved genes
across species in development. Furthermore, we applied ManiNetCluster
to time series transcriptome data measured in the green alga
Chlamydomonas reinhardtii to discover the genomic functions linking
various metabolic processes between the light and dark periods of a
diurnally cycling culture;



(2) Varmole (Nguyen, et al., Bioinformatics, 2020) is an interpretable
deep learning method that simultaneously reveals genomic functions and
mechanisms while predicting phenotype from genotype. In particular,
Varmole embeds multi-omic networks into a deep neural network
architecture and prioritizes variants, genes and regulatory linkages
via biological drop-connect without needing prior feature selections.
With an application to schizophonia, we demonstrate that Varmole
provides an effective alternative for recent statistical methods that
associate functional omic data (e.g. gene expression) with genotype
and phenotype and that link variants to individual genes in population
studies such as genome-wide association study;



(3) ECMarker (Jin*, Nguyen*, et al., Bioinformatics, 2020) is an
interpretable and scalable machine learning model that predicts gene
expression biomarkers for disease phenotypes and simultaneously
reveals underlying regulatory mechanisms. Particularly, ECMarker is
built on the integration of semi- and discriminative- restricted
Boltzmann machines, a neural network model for classification allowing
lateral connections at the input gene layer. With application to the
gene expression data of non-small cell lung cancer (NSCLC) patients,
we found that ECMarker not only achieved a relatively high accuracy
for predicting cancer stages but also identified the biomarker genes
and gene networks implying the regulatory mechanisms in lung cancer
development.



Finally, we propose a novel multiview learning method, Malignomics, to
predict phenotypes from heterogeneous multi-omic features. Malignomics
will first align multi-omic features by deep manifold alignment onto a
common latent space, better predicting nonlinear relationships across
omics. This deep alignment aims to preserve both global consistency
and local smoothness across omics and reveal higher-order nonlinear
interactions (i.e., manifolds) among cross-omic features. Second, it
uses these manifold structures to regularize the classifiers for
predicting phenotypes. This manifold-regularization allows
highlighting cross-omic feature manifolds and prioritizing the
features and interactions for the phenotypes. The prioritized
multi-omic features will further reveal underlying phenotypic
functions and mechanisms and thus enhance the biological
interpretation of Malignomics. We will apply Malignomics to
multi-omics data in neuropsychiatric disorders, and prioritize gene
regulatory networks linking risk variants, regulatory elements, and
genes for the disorders. We will also compare Malignomics with the
state-of-the-arts, and investigate how the manifold regulation will
potentially improve understanding of multi-omics functions and
predicting diseases.

CSE 600 Seminar Series | Fall 2025

Speaker: Prof. R Sekar