A5007489034- Advanced Graph Neural Networks
- Explainable Artificial Intelligence (XAI)
- Topic Modeling
- Adversarial Robustness in Machine Learning
- Anomaly Detection Techniques and Applications
- Recommender Systems and Techniques
- Natural Language Processing Techniques
- Complex Network Analysis Techniques
- Network Security and Intrusion Detection
- Machine Learning in Healthcare
- Advanced Malware Detection Techniques
- Multimodal Machine Learning Applications
- Ethics and Social Impacts of AI
- Machine Learning and Data Classification
- Artificial Intelligence in Healthcare and Education
- Machine Learning in Materials Science
- Data Stream Mining Techniques
- Semantic Web and Ontologies
- Bioinformatics and Genomic Networks
- Radiomics and Machine Learning in Medical Imaging
- Advanced Neural Network Applications
- Domain Adaptation and Few-Shot Learning
- Advanced Text Analysis Techniques
- Bayesian Modeling and Causal Inference
- Graph Theory and Algorithms
University of Georgia
2022-2025
Baotou Medical College
2024
Texas A&M University
2016-2022
Mitchell Institute
2021
South China University of Technology
2014
Uncovering the mysterious ways machine learning models make decisions.
Large language models (LLMs) have demonstrated impressive capabilities in natural processing. However, their internal mechanisms are still unclear and this lack of transparency poses unwanted risks for downstream applications. Therefore, understanding explaining these is crucial elucidating behaviors, limitations, social impacts. In article, we introduce a taxonomy explainability techniques provide structured overview methods Transformer-based models. We categorize based on the training...
Graph Neural Networks (GNNs) have shown superior performance in analyzing attributed networks various web-based applications such as social recommendation and web search. Nevertheless, high-stake decision-making scenarios online fraud detection, there is an increasing societal concern that GNNs could make discriminatory decisions towards certain demographic groups. Despite recent explorations on fair GNNs, these works are tailored for a specific GNN model. However, myriads of variants been...
Text data augmentation is an effective strategy for overcoming the challenge of limited sample sizes in many natural language processing (NLP) tasks. This especially prominent few-shot learning scenario, where target domain generally much scarcer and lowered quality. A widely-used to mitigate such challenges perform better capture invariance increase size. However, current text methods either can't ensure correct labeling generated (lacking faithfulness) or sufficient diversity compactness),...
Self-supervised learning (SSL) has been demonstrated to be effective in pre-training models that can generalized various downstream tasks. Graph Autoencoder (GAE), an increasingly popular SSL approach on graphs, widely explored learn node representations without ground-truth labels. However, recent studies show existing GAE methods could only perform well link prediction tasks, while their performance classification tasks is rather limited. This limitation casts doubt the generalizability...
This study investigates the application of large language models (LLMs), specifically GPT-3.5 and GPT-4, with Chain-of-Though (CoT) in automatic scoring student-written responses to science assessments. We focused on overcoming challenges accessibility, technical complexity, lack explainability that have previously limited use artificial intelligence-based tools among researchers educators. With a testing dataset comprising six assessment tasks (three binomial three trinomial) 1650 student...
Large pre-trained models, also known as foundation models (FMs), are trained in a task-agnostic manner on large-scale data and can be adapted to wide range of downstream tasks by fine-tuning, few-shot, or even zero-shot learning. Despite their successes language vision tasks, we have not yet seen an attempt develop for geospatial artificial intelligence (GeoAI). In this work, explore the promises challenges developing multimodal GeoAI. We first investigate potential many existing FMs testing...
With the widespread use of deep neural networks (DNNs) in high-stake applications, security problem DNN models has received extensive attention. In this paper, we investigate a specific called trojan attack, which aims to attack deployed systems relying on hidden trigger patterns inserted by malicious hackers. We propose training-free approach is different from previous work, trojaned behaviors are injected retraining model poisoned dataset. Specifically, do not change parameters original...
While deep neural networks (DNN) have become an effective computational tool, the prediction results are often criticized by lack of interpretability, which is essential in many real-world applications such as health informatics. Existing attempts based on local interpretations aim to identify relevant features contributing most DNN monitoring neighborhood a given input. They usually simply ignore intermediate layers that might contain rich information for interpretation. To bridge gap, this...
Recent methods in sequential recommendation focus on learning an overall embedding vector from a user's behavior sequence for the next-item recommendation. However, empirical analysis, we discovered that often contains multiple conceptually distinct items, while unified is primarily affected by one's most recent frequent actions. Thus, it may fail to infer next preferred item if similar items are not dominant interactions. To this end, alternative solution represent each user with vectors...
Recommender systems in industry generally include two stages: recall and ranking. Recall refers to efficiently identify hundreds of candidate items that user may interest from a large volume item corpus, while the latter aims output precise ranking list using complex models. Recently, graph representation learning has attracted much attention supporting high quality search at scale. Despite its effectiveness embedding vectors for objects user-item interaction network, computational costs...
Machine learning models are becoming pervasive in high-stakes applications. Despite their clear benefits terms of performance, the could show discrimination against minority groups and result fairness issues a decision-making process, leading to severe negative impacts on individuals society. In recent years, various techniques have been developed mitigate unfairness for machine models. Among them, in-processing methods drawn increasing attention from community, where is directly taken into...
Graph neural networks (GNNs) have received remarkable success in link prediction (GNNLP) tasks. Existing efforts first predefine the subgraph for whole dataset and then apply GNNs to encode edge representations by leveraging neighborhood structure induced fixed subgraph. The prominence of GNNLP methods significantly relies on adhoc Since node connectivity real-world graphs is complex, one shared limited all edges. Thus, choices subgraphs should be personalized different However, performing...
Artificial general intelligence (AGI) has gained global recognition as a future technology due to the emergence of breakthrough large language models and chatbots such GPT-4 ChatGPT, respectively. Compared conventional AI models, typically designed for limited range tasks, demand significant amounts domain-specific data training may not always consider intricate interpersonal dynamics in education. AGI, driven by recent pre-trained represents leap capability machines perform tasks that...
At the dawn of Artificial General Intelligence (AGI), emergence large language models such as ChatGPT show promise in revolutionizing healthcare by improving patient care, expanding medical access, and optimizing clinical processes. However, their integration into systems requires careful consideration potential risks, inaccurate advice, privacy violations, creation falsified documents or images, overreliance on AGI education, perpetuation biases. It is crucial to implement proper oversight...
Deep Learning has been successfully applied to many application domains, yet its advantages have slow emerge for time series forecasting. For example, in the well-known Makridakis (M) Competitions, hybrids of traditional statistical or machine learning techniques only recently become top performers. With recent architectural advances deep being forecasting (e.g., encoder-decoders with attention, transformers, and graph neural networks), begun show significant advantages. Still, area pandemic...
Sequential recommendation has become increasingly essential in various online services. It aims to model the dynamic preferences of users from their historical interactions and predict next items. The accumulated user behavior records on real systems could be very long. This rich data brings opportunities track actual interests users. Prior efforts mainly focus making recommendations based relatively recent behaviors. However, overall sequential may not effectively utilized, as early might...
This work develops \emph{mixup for graph data}. Mixup has shown superiority in improving the generalization and robustness of neural networks by interpolating features labels between two random samples. Traditionally, can on regular, grid-like, Euclidean data such as image or tabular data. However, it is challenging to directly adopt augment because different graphs typically: 1) have numbers nodes; 2) are not readily aligned; 3) unique typologies non-Euclidean space. To this end, we propose...
Graph Neural Networks (GNNs) have emerged as the leading paradigm for solving graph analytical problems in various real-world applications. Nevertheless, GNNs could potentially render biased predictions towards certain demographic subgroups. Understanding how bias arises is critical, it guides design of GNN debiasing mechanisms. However, most existing works overwhelmingly focus on debiasing, but fall short explaining such induced. In this paper, we study a novel problem interpreting...
Machine learning (ML) systems have been increasingly applied in web security applications such as spammer detection, malware detection and fraud detection. These an intrinsic adversarial nature where intelligent attackers can adaptively change their behaviors to avoid being detected by the deployed detectors. Existing efforts against adversaries are usually limited type of ML models or specific image classification. Additionally, working mechanisms cannot be well understood users, which turn...
While outlier detection has been intensively studied in many applications, interpretation is becoming increasingly important to help people trust and evaluate the developed models through providing intrinsic reasons why given outliers are identified. It a nontrivial task for interpreting abnormality of due distinct characteristics different models, complicated structures data certain imbalanced distribution normal instances. In addition, contexts where locate, as well relation between...