Nghia Duong Tan

Dr. Nghia Duong Tan

School of Electronics and Telecommunications, Hanoi University of Science and Technology
orcid

Research Focus & Metrics

Academic Career Length

14Years (2013-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
21 April 2020 in IEEE Access
Verified
Journal Article
Open access
Nghia Duong TanNghia Duong Tan
Nghia Duong Tan
TVTuan Anh Vuong
Tuan Anh Vuong
DNDuc Minh Nguyen
Duc Minh Nguyen
QDQuang Hieu Dang
Quang Hieu Dang
While collaborative filtering (CF) is the most popular approach for recommendation systems, it only makes use of the ratings given to items by users and neglects side information about user attributes or item features. In this work, a natural language processing (NLP) technique is applied to generate a more consistent version of Tag Genome, a side information which is associated with each movie in the MovieLens 20M dataset. Subsequently, we propose a 3-layer autoencoder to create a more compact representation of these tags which improves the performance of the system both in accuracy and in computational complexity. Finally, the proposed representation and the well-known matrix factorization techniques are combined into a unified framework that outperforms the state-of-the-art models by at least 2.87% and 3.36% in terms of RMSE and MAE, respectively.

1 November 2018
Verified
Conference Paper
Closed access
Nghia Duong TanNghia Duong Tan
Nghia Duong Tan
VTViet Duc Than
Viet Duc Than
TTTrong Hiep Tran
Trong Hiep Tran
QDQuang Hieu Dang
Quang Hieu Dang
DNDuc Minh Nguyen
Duc Minh Nguyen
HPHung Manh Pham
Hung Manh Pham
Thanks to its successful application in recommendation systems, collaborative filtering (CF) technique has become one of the most popular research topic in data mining and information retrieval. The two more dominant approaches to CF are neighborhood-based models and latent factor models. Currently, the state-of-the-art recommendation systems are mainly based on the latter approach owing to its ability to explore hidden factors connecting items to users. In this work, we propose a novel similarity measure which helps improve the accuracy of the neighborhood-based models. Intensive experiments show that our proposed model not only outperforms the best latent factor model with respect to RMSE by 2.2% but also works at least 2 times faster than its counterpart.

1 June 2017
Verified
Conference Paper
Closed access
MTManh Hoang Tran
Manh Hoang Tran
Nghia Duong TanNghia Duong Tan
Nghia Duong Tan
DNDuc Minh Nguyen
Duc Minh Nguyen
QDQuang Hieu Dang
Quang Hieu Dang
In this paper, we introduce a hybrid fingerprint matcher which consists of two stages: local minutiae matching stage and consolidation stage. In the former stage, each minutia in the fingerprint is represented by a local feature vector which is built from the features of the neighboring minutiae within a fixed-radius circle around it. The features of each neighboring minutia are defined according to the spatial relationship with the central minutia and used for matching task. New calculation methods of the adaptive matching conditions for two neighboring minutiae and the local similarity score between two central minutiae are introduced. The latter stage performs multiple alignments based on the pairs of matching minutiae to produce the final matching score between two fingerprints. Extensive experiments over a public database used in Fingerprint Verification Competition 2002 (FVC2002) prove a considerable improvement in accuracy of our proposed algorithm against some well-known techniques and Top 5 participants of FVC2002.

4 January 2022 in Future Internet
Verified
Journal Article
Open access
PDF file
Nghia Duong TanNghia Duong Tan
Nghia Duong Tan
Nam Nguyen DoanNam Nguyen Doan
Nam Nguyen Doan
Truong Giang DoTruong Giang Do
Truong Giang Do
MTManh Hoang Tran
Manh Hoang Tran
MNMinh Nguyen
Minh Nguyen
QDQuang Hieu Dang
Quang Hieu Dang
Recommendation systems based on convolutional neural network (CNN) have attracted great attention due to their effectiveness in processing unstructured data such as images or audio. However, a huge amount of raw data produced by data crawling and digital transformation is structured, which makes it difficult to utilize the advantages of CNN. This paper introduces a novel autoencoder, named Half Convolutional Autoencoder, which adopts convolutional layers to discover the high-order correlation between structured features in the form of Tag Genome, the side information associated with each movie in the MovieLens 20 M dataset, in order to generate a robust feature vector. Subsequently, these new movie representations, along with the introduction of users’ characteristics generated via Tag Genome and their past transactions, are applied into well-known matrix factorization models to resolve the initialization problem and enhance the predicting results. This method not only outperforms traditional matrix factorization techniques by at least 5.35% in terms of accuracy but also stabilizes the training process and guarantees faster convergence.

15 August 2019
Verified
Book Chapter
Closed access
Nghia Duong TanNghia Duong Tan
Nghia Duong Tan
VTViet Duc Than
Viet Duc Than
TVTuan Anh Vuong
Tuan Anh Vuong
TTTrong Hiep Tran
Trong Hiep Tran
QDQuang Hieu Dang
Quang Hieu Dang
HPHung Manh Pham
Hung Manh Pham

1 November 2018
Verified
Conference Paper
Closed access
Nghia Duong TanNghia Duong Tan
Nghia Duong Tan
VTViet Duc Than
Viet Duc Than
TTTrong Hiep Tran
Trong Hiep Tran
TPThi Hong Anh Pham
Thi Hong Anh Pham
VNVan Hoang Anh Nguyen
Van Hoang Anh Nguyen
NTNam Hoang Tran
Nam Hoang Tran
Cold-start problem, occurring when a new user joins the system, is an important factor that influences the satisfaction of the users. Since the user has no or almost no interaction with the recommendation system before, the system finds it difficult to obtain adequate information about user preferences so that it cannot provide a high quality personalized recommendation service. Furthermore, there is also a lack of information about the preferences of even familiar users due to the fact that many of them are not always willing to explicitly describe their evaluation of a specific item through ratings. Utilizing all information about user preferences including explicit and especially implicit data helps our recommendation system achieve a promising result in the ACM RecSys Challenge 2018 organized by Spotify. Experiments show that the proposed model not only deals with the cold-start problem but also gains a high precision of recommendation for the whole system whilst costing an amount of much lower time and hardware resource compared with Top 5 participants.