蒙特卡洛的Faith

About Me

Hey, there! I am 程序员 Faith. I develop novel machine learning solutions for Software Engineering problems. I am also broadly interested in leveraging AI to help people become more productive in professional and everyday life.

My research interest has been under the following themes:

AI for Software Engineering
Unsupervised Learning

💪 My programming skills:

I love

but I need a life-long learning on it. I am the fans of Mac but I am also found of using

as well as

.

Project Highlights

Micronnx

A Deep Learning inference framework in Pure C code without any dependency.

Aicmder

Aicmder is a quick command-line solution for Deep Learning application. Installed by pip install aicmder (Can work with label-studio for auto labeling)

Docker image

mydocker is a collections of docker images including cuda, cuda with torch images and Klipper for 3D printing. Jupyterhub is a jupyterhub docker image combined with autograder for teaching purpose. octoprint-docker is a octoprint docker image with some customize configuration.

Publications

2022

Bounded Asymmetric Gaussian Mixture-Based Hidden Markov Models
Z Xian, M Azam, M Amayri, W Fan, N Bouguila
Hidden Markov Models and Applications. Unsupervised and Semi-Supervised Learning. Springer, Cham
Not available Abstract Bibtex

Hidden Markov models (HMMs) have been widely applied in machine learning to model diversified and heterogeneous time series data. In this chapter, integration of the bounded asymmetric Gaussian mixture model (BAGMM) into the framework of HMM is proposed, which uses BAGMM to model the emission probabilities, which are also known as observation probabilities. The transition probabilities of HMM must be discrete, but emission probabilities can be modeled by any continuous distribution. This modified HMM (BAGMM-HMM) is motivated by the proven capability of BAGMM and its benefits over Gaussian mixture models (GMMs). The complete inference and parameter learning of BAGMM-HMM are proposed. The validation involves several real-world applications, such as occupancy estimation and human activity recognition. The proposed model is compared with three existing mixture model-based HMMs in a similar setting for all the applications. It is examined from all experiments and results that BAGMM-HMM has good modeling capabilities and outperforms other comparable mixture-based HMMs.

@incollection{xian2022bounded,
   title={Bounded Asymmetric Gaussian Mixture-Based Hidden Markov Models},
   author={Xian, Zixiang and Azam, Muhammad and Amayri, Manar and Fan, Wentao and Bouguila, Nizar},
   booktitle={Hidden Markov Models and Applications},
   pages={33--58},
   year={2022},
   publisher={Springer}
}

2021

Statistical modeling using bounded asymmetric gaussian mixtures application to human action and gender recognition
Z Xian, M Azam, N Bouguila
2021 IEEE 22nd International Conference on Information Reuse and Integration for Data Science (IRI)
PDF Abstract Bibtex

To determine the structure of high dimensional data without knowing the number of clusters nor the importance of the involved features, we propose an unsupervised feature selection framework using the bounded asymmetric Gaussian mixture model (BAGMM-FS). The bounded asymmetric Gaussian distribution has an asymmetric shape and bounded range, making it a good choice for modeling real-world data. We propose a parameter learning approach based on the expectation-maximization (EM) algorithm, and we approach the model selection task using the minimum message length (MML) criterion. The validation involves several human-related recognition challenges, such as human activity categorization and human gender recognition. It's examined from all experiments and results that BAGMM-FS has good modeling capabilities and outperforms other comparable mixture models, especially for high dimensional complex datasets.

@inproceedings{xian2021statistical,
    title={Statistical Modeling Using Bounded Asymmetric Gaussian Mixtures: Application to Human Action and Gender Recognition},
    author={Xian, Zixiang and Azam, Muhammad and Bouguila, Nizar},
    booktitle={2021 IEEE 22nd International Conference on Information Reuse and Integration for Data Science (IRI)},
    pages={41--48},
    year={2021},
    organization={IEEE}
}

Model Selection Criterion for Multivariate Bounded Asymmetric Gaussian Mixture Model
Z Xian, M Azam, M Amayri, N Bouguila
2021 29th European Signal Processing Conference (EUSIPCO)
PDF Abstract Bibtex

In this paper, model selection criterion for bounded support asymmetric Gaussian mixture model (BAGMM) using minimum message length (MML) is proposed. The proposed approach is validated using synthetic data, real data and occupancy detection application. The proposed approach is compared with other state of the art model selection approaches. Moreover, the developed bounded mixture is compared with asymmetric Gaussian mixture model (AGMM).

@inproceedings{xian2021model,
   title={Model Selection Criterion for Multivariate Bounded Asymmetric Gaussian Mixture Model},
   author={Xian, Zixiang and Azam, Muhammad and Amayri, Manar and Bouguila, Nizar},
   booktitle={2021 29th European Signal Processing Conference (EUSIPCO)},
   pages={1436--1440},
   year={2021},
   organization={IEEE}
}

Miscellaneous

I became a PhD student at Macau University of Science and Technology.

I received my master degree from Concordia University.

I became a master student at Concordia University.

Faith 程序员 文章列表new