A5102924905- Recommender Systems and Techniques
- Energy Load and Power Forecasting
- Sentiment Analysis and Opinion Mining
- Tensor decomposition and applications
- Hydrological Forecasting Using AI
- Spam and Phishing Detection
- Face and Expression Recognition
- Internet Traffic Analysis and Secure E-voting
- Advanced Text Analysis Techniques
- Network Security and Intrusion Detection
- Web Data Mining and Analysis
- Advanced Graph Neural Networks
- Topic Modeling
- Data Mining Algorithms and Applications
- Domain Adaptation and Few-Shot Learning
- Generative Adversarial Networks and Image Synthesis
- Image Retrieval and Classification Techniques
- Artificial Intelligence in Healthcare
- Statistical Methods in Epidemiology
- Machine Learning in Healthcare
- Human Mobility and Location-Based Analysis
- Calcium Carbonate Crystallization and Inhibition
- Advanced Image and Video Retrieval Techniques
- Asphalt Pavement Performance Evaluation
- Corrosion Behavior and Inhibition
Walmart (United States)
2021-2024
China Southern Power Grid (China)
2021
Michigan State University
2014-2019
Samsung (United States)
2015
Putian University
2014
Harbin Institute of Technology
2009
North Carolina State University
1994
One major limitation of linear models is the lack capability to capture predictive information from interactions between features. While introducing high-order feature interaction terms can overcome this limitation, approach tremendously increases model complexity and imposes significant challenges in learning against overfitting. In paper, we proposed a novel Multi-Task Interaction Learning~(MTIL) framework exploit task relatedness interactions, which provides better generalization...
Medical predictive modeling is a challenging problem due to the heterogeneous nature of patients. In order build effective medical models we need address such during and allow patients have their own personalized instead using one-size-fits-all model. However, building model for each patient computationally expensive over-parametrization makes it susceptible overfitting problem. To these challenges, propose novel approach called FactORized MUlti-task LeArning (FORMULA), which learns via...
Ensemble forecasting is a widely-used numerical prediction method for modeling the evolution of nonlinear dynamic systems. To predict future state such systems, set ensemble member forecasts generated from multiple runs computer models, where each run obtained by perturbing starting condition or using different model representation system. The mean median typically chosen as point estimate forecasts. These approaches are limited in that they assume equally skillful and may not preserve...
Previous chapter Next Full AccessProceedings Proceedings of the 2016 SIAM International Conference on Data Mining (SDM)GSpartan: a Geospatio-Temporal Multi-task Learning Framework for Multi-location PredictionJianpeng Xu, Pang-Ning Tan, Lifeng Luo, and Jiayu ZhouJianpeng Zhoupp.657 - 665Chapter DOI:https://doi.org/10.1137/1.9781611974348.74PDFBibTexSections ToolsAdd to favoritesExport CitationTrack CitationsEmail SectionsAboutAbstract This paper presents novel geospatio-temporal prediction...
The rapid evolution of text-to-image diffusion models has opened the door generative AI, enabling translation textual descriptions into visually compelling images with remarkable quality. However, a persistent challenge within this domain is optimization prompts to effectively convey abstract concepts concrete objects. For example, text encoders can hardly express "peace", while easily illustrate olive branches and white doves. This paper introduces novel approach named Prompt Optimizer for...
Ensemble forecasting is a well-known numerical prediction technique for modeling the evolution of nonlinear dynamic systems. The ensemble member forecasts are generated from multiple runs computer model, where each run obtained by perturbing starting condition or using different model representation system. mean median typically chosen as consensus point estimate aggregated decision making purposes. These approaches limited in that they assume equally skill ful and do not consider their...
Predictive modeling of large-scale spatio-temporal data is an important but challenging problem as it requires training models that can simultaneously predict the target variables interest at multiple locations while preserving spatial and temporal dependencies data. In this paper, we investigate effectiveness applying a multi-task learning approach based on supervised tensor decomposition to prediction problem. Our proposed framework, known SMART, encodes third-order extracts set...
Previous chapter Next Full AccessProceedings Proceedings of the 2016 SIAM International Conference on Data Mining (SDM)Synergies that Matter: Efficient Interaction Selection via Sparse Factorization MachineJianpeng Xu, Kaixiang Lin, Pang-Ning Tan, and Jiayu ZhouJianpeng Zhoupp.108 - 116Chapter DOI:https://doi.org/10.1137/1.9781611974348.13PDFBibTexSections ToolsAdd to favoritesExport CitationTrack CitationsEmail SectionsAboutAbstract Collaborative filtering has been widely used in modern...
This paper considers an approach to identify previously undetected malicious clients in Internet Service Provider (ISP) networks by combining flow classification with a graph-based score propagation method. Our represents all HTTP communications between and servers as weighted, near-bipartite graph, where the nodes correspond IP addresses of while links are their interconnections, weighted according output flow-based classifier. We employ two-phase alternating algorithm on graph suspicious...
This paper considers an approach to identify previously undetected malicious clients in Internet Service Provider (ISP) networks by combining flow classification with a graph-based score propagation method. Our represents all HTTP communications between and servers as weighted, near-bipartite graph, where the nodes correspond IP addresses of while links are their interconnections, weighted according output flow-based classifier. We employ two-phase alternating algorithm on graph suspicious...
In climate and environmental sciences, vast amount of spatio-temporal data have been generated at varying spatial resolutions from satellite observations computer models. Integrating such diverse sources has proven to be useful for building prediction models as the multi-scale may capture different aspects Earth system. this paper, we present a novel framework called MUSCAT predictive modeling multi-scale, data. performs joint decomposition multiple tensors scales, taking into account...
Abstract An in‐situ FTIR–ATR method has been used to monitor the sorption processes of water and pH 1.3‐sulfurous acid in two latex paints base polymer common both. The kinetics could not be described by a simple Fickian model. spectra also showed evidence swelling, which was confirmed separate swelling measurements. Anomalous behavior noted for paint containing CaCO 3 when exposed sulfurous acid. amount sorbed this sample went through maximum, then decreased constant level. This accompanied...
This paper presents a novel multi-task learning framework for the accurate prediction of spatio-temporal data at multiple locations. The encodes as third-order tensor and performs supervised decomposition to identify latent factors that capture inherent spatiotemporal variabilities their relationship target variable interest. is unique in it trains both spatial temporal models from decomposed aggregates outputs generate its final prediction. model parameters are simultaneously estimated by...
Recommender systems are popular tools for information retrieval tasks on a large variety of web applications and personalized products. In this work, we propose Generative Adversarial Network based recommendation framework using positive-unlabeled sampling strategy. Specifically, utilize the generator to learn continuous distribution user-item tuples design discriminator be binary classifier that outputs relevance score between each user item. Meanwhile, is applied in learning procedure...
Frequent closed pattern mining has been developed for decades, mostly on a two dimensional matrix. This paper addresses the problem of high frequent patterns (nFCPs) from dense binary dataset, where dataset is represented by cube. As existing FP-tree or enumeration tree based algorithms do not suit n-dimensional data, we are motivated to propose novel algorithm called HDminer nFCPs mining. employs effective search space partition and pruning strategies enhance efficiency. We have implemented...
Similar item recommendation is a critical task in the e-Commerce industry, which helps customers explore similar and relevant alternatives based on their interested products. Despite traditional machine learning models, Graph Neural Networks (GNNs), by design, can understand complex relations like similarity between However, contrast to wide usage retrieval tasks focus optimizing relevance, current GNN architectures are not tailored toward maximizing revenue-related objectives such as Gross...
Feature selection has been studied widely in literatures supervised learning scenarios. methods are categorized into two classes: ¿wrapper¿ and ¿filter¿ approaches. In this paper, we propose a filter method called cloud score based on membership model fuzzy field. We determine the discrimination power of certain feature by to evaluate feature's importance. This is compared with variance Fisher UCI iris dataset. The results experiments demonstrate feasibility our algorithm.
Recommendation systems are used widely across many industries, such as e-commerce, multimedia content platforms and social networks, to provide suggestions that a user will most likely consume or connect; thus, improving the experience. This motivates people in both industry research organizations focus on personalization recommendation algorithms, which has resulted plethora of papers. While academic mostly focuses performance algorithms terms ranking quality accuracy, it often neglects key...