Hey There
I AM DUONG CONG SON
I'm currently majoring in Computer Science.
duongcongson01@gmail.com
0981902001
Education
Student at Hanoi University of Civil Engineering (HUCE)
Major: Computer Science
GPA: 3.72/4.0
Selected Awards/Honors
• CSC Award top 10 excellent students selected by Hanoi University of Civil Engineering, 2022 • Do Quoc Sam scholarship award top 20 students with the highest GPA, 2022 • Best Poster Award The International Symposium on Information and Communication Technology, SoICT 2022 • Second Prize, Scientific Research Competition for HUCE Students, 2022 Applying Machine Learning to Software-defined Network for intelligent routing. • Second Prize, Scientific Research Competition for HUCE Students, 2022 Transformer-based machine translation preference with a bilingual Chatbot. • First Prize, Scientific Research Competition for HUCE Students, 2021 Replication and consistency in heterogeneous and Distributed SDN networks
About me
Hello, I’m Cong Son, Computer Science student at Ha Noi University of Civil Engineering. * Able to work under pressure, easy to adapt to the working environment and conditions. * Interested in developing practical solutions to real-world problems and designing/constructing products that enhance human experiences. * Expertise in high-dimensional statistics, statistical models, machine learning, and deep learning, as well as hands-on experience.
My Publications
[Paper] Server and Route Selection Optimization for Knowledge-defined Distributed Network Based on Gambling Theory and LSTM Neural Networks
This paper focuses primarily on the server selection mechanism, aiming to improve network performance, such as bandwidth, frequency spectrum, and power. These parameters are used as objective functions for the proposed algorithm.
Paper: Globecom conference (IEEE Global Communications Conference) 2023 in Malaysia.
[Paper] LSTM-BASED APPROACH FOR SERVER AND ROUTE SELECTION IN INTER-SDN DOMAINS
We propose a link cost prediction mechanism using Long Short-Term Memory Network (LSTM) to optimize our server and route selection. The objective of using LSTM is to learn character- istics of non-linear nature and uncertainty of traffic flows. It can take advantage of historical traffic parameters instead of considering only current network states. The predicted link cost consists of several network parameters, such as packet loss, link utilization, delay, and link overhead
Paper: Journal of Computer Science and Cybernetics.
[Paper] Knowledge-defined Heterogeneous Network: Use-case of QoS-based Server and Route Selection in Large-scale Network
We present a QoS-based server and route selection mechanism to select server and routing paths with optimized cost in order to optimize network performance. The experiments show that our proposal provides a good trade-off between QoS metrics (improvements of 15 percent of link utilization 13 percent of loss) and greatly reduced overhead and response time, compared with benchmarks.
Paper: International Symposium on Information and Communication Technology (SoICT 2022)
My Projects
[PERSIONAL] Hệ thống CRM và Website thiết kế
Homepage comprises features (e.g., the ability to view a flip book and various animation effects...). The CRM page facilitates tasks (e.g., salary calculation, project management, and personnel administration...).
Design UX/UI and implement additional functionalities (both back-end and front-end) based on
the requirements from stakeholders. Design database schema and implement back-end services. Collaborate with team
leader to define and improve the team’s workflow, which decreases 40% time to deploy new functionality and reduce bugs
related to configuration
Demo font-end: https://thietkehaco.com/
Technologies: Python, Flask API, MongoDB, React.js, TypeScript, CI/CD, Docker
[PERSIONAL] SDN Inter-domains network application for a heterogeneous and distribute SDN network
Built an east-west interface to ensure data consistency between many heterogeneous SDN controllers.
Applied the quorum-based replication mechanism using the Q-Learning algorithm which could guarantee an acceptable consistency level while maintaining good network performance metrics, such as read delay, write delay and overhead.
Concretely, the three cases of read thresholds 10ms, 30ms, and 50ms had convergent mean values of 0.4, 0.11, and 0.3, respectively. And version staleness metric was always less than the fixed version staleness threshold (0.5).
Technologies: Java, Maven, Python, MongoDB
[HUCE] Transformer-based machine translation preference with a bilingual Chatbot
Proposed transformer algorithm in machine translation with collected data and preprocessed datasets such as removing HTML tags, stop words, word separation, word embedding with over 120 thousand of sentences.
The BLEU score of the proposed method ranked in the first position at 19.1, followed by Seq2Seq with attention (16.4) and Seq2seq (4.3). Furthermore, training time was 2x faster than others methods.
Built and deployed an online bilingual chatbot with Flask and TensorFlow, allowing users to ask questions in English or Vietnamese, responded in the appropriate language.
Technologies: python, Tensorflow, NLTK, Flask
[HuceInTech] An active and intelligent network management system by voice
Prevent network congestion: Routing algorithm, Traffic Classification, Server Selection
Data security and privacy: Blockchain
Application: React Native, Flask
Technologies: python, react native, tensorflow
[HUCE] Information Recognition on the University Test Paper
Image Pre-processing using OpenCV cuts the time it takes to use the EAST technique in half while maintaining the same level of accuracy
Generate multi-digit numbers from the MNIST dataset with the right Data Augmentation methods to help increase the performance to 40%
The combination of CRNN-CTC Loss Model, Attention Layer, Lexicon Search improve the outcome significantly
Final accuracy of 97.14% on a class list of 245 students, inference time around 0.9–1s/image
Technologies: Python, OpenCV, PyImage
[HUCE] Logistic Regression and KNN on MNIST dataset from scratch
Implement and train a logistic regression model from scratch in Python on the MNIST dataset (no Library). The logistic regression model should be trained on the Training Set using stochastic gradient descent. It should achieve 90-93% accuracy on the Test Set.
Technologies: Python
[Persional] Website My study space
Build a website, help everyone can focus on studying by listening to relax music and break time. Features such as listening to music and to do list are integrated in the website (no Library).
Technologies: HTML, CSS, JavaScript
[Persional] Breast Cancer Prediction
Predicting whether cancer is benign or malignant using Logistic Regression (Binary Class Classification)
Dataset Used: Breast Cancer Wisconsin (Diagnostic) Dataset.
Accuracy of 91.95 % (Training Data) and 91.81 % (Test Data).
Technologies: Python
[Persional] Learning Data Structures and Algorithms (for education)
Data structures allow you to organize data in a particular way efficiently. They are critical to any problem, provide a complete solution, and act like reusable code.
Technologies: Python
[Persional] Compare divide-and-conquer algorithm with traditional algorithms-O(n^2) (for education)
Visualize divide-and-conquer algorithms with traditional algorithms-O(n^2) in JavaScript, I have successfully completed the class assignment assigned in the university. Help me better understand how each algorithm works...
Technologies: HTML, CSS, JavaScript
[HUCE] Database system project
Đồ án thiết kế cơ sở dữ liệu, xây dựng hệ thống logistic phân tích thiết kế hệ cơ sở dữ liệu chặt chẽ có sử dụng trigger, stored procedure, business rule. Đạt điểm tuyệt đối trong lần bảo vệ.
Technologies: php, sql
[HUCE] Assignment machine learning ( kmean, Naive Bayes )
Naive Bayes: Handwriting recognition is the ability of a computer to interpret hand written text as the characters. In this assignment we will be trying to recognize numbers from images. To accomplish this task, we will be using a Naive Bayes classifier.
K-mean: Customer Segmentation can be a powerful means to identify unsatisfied customer needs. This technique can be used by companies to outperform the competition by developing uniquely appealing products and services.
Technologies: python
[Persional] Recommender Products
https://github.com/CongSon01/RecommenderProducts
Technologies: python