Trainee Project
Title:
Characteristics of MRDM algorithms: A comparison between RCA, PCA, and Graph-FCA
Dates:
2022/11/15 - 2023/05/15
Supervisor(s):
Description:
Multi-relational Data Mining (MRDM) [1] is the process of discovering knowledge or patterns from massive amounts of data (data
mining), when the data comes from heterogeneous linked sources (multi-relational). Additionally, unsupervised learning is the name
given to the process of extracting patterns from unlabeled data. Several mathematical frameworks have been proposed to deal with
this task, having their strengths and weaknesses each.
In this work, we aim to study from a computer science point of view, the characteristics of the frameworks of Relational Concept
Analysis (RCA), Polyadic Concept Analysis (PCA), and Graph Formal Concept Analysis (G-FCA). Particularly, we are interested in
putting together the differences in their main algorithms in terms of temporal and spatial complexities, practical efficiency in different
scenarios and finally, the differences in their outputs.
The objective of this study is to provide insights on when and why it is convenient to use any of the approaches. In addition, to
discuss the limitations of the implemented algorithms and provide suggestions on the possibilities of improvement. Furthermore, we
aim to provide an open source implementation of the frameworks.
The master thesis project would include the following stages:
1. Collect the required bibliography and understand the basic theoretical context in each framework.
2. Implement at least one algorithm of each framework using the same programming language so that cross-experimentation makes
sense.
3. Design and propose experiments to get interesting comparison between the implementations.
4. Draw a conclusion out of the study.
References
[1] Saso Dzeroski. "Multi-relational data mining: an introduction". In: ACM SIGKDD Explorations Newsletter 5.1 (July 2003), pp.
1-16. issn: 1931-0145. doi: 10.1145/959242.959245. url: https://doi.org/10.1145/959242.959245 (visited on 04/04/2022).
mining), when the data comes from heterogeneous linked sources (multi-relational). Additionally, unsupervised learning is the name
given to the process of extracting patterns from unlabeled data. Several mathematical frameworks have been proposed to deal with
this task, having their strengths and weaknesses each.
In this work, we aim to study from a computer science point of view, the characteristics of the frameworks of Relational Concept
Analysis (RCA), Polyadic Concept Analysis (PCA), and Graph Formal Concept Analysis (G-FCA). Particularly, we are interested in
putting together the differences in their main algorithms in terms of temporal and spatial complexities, practical efficiency in different
scenarios and finally, the differences in their outputs.
The objective of this study is to provide insights on when and why it is convenient to use any of the approaches. In addition, to
discuss the limitations of the implemented algorithms and provide suggestions on the possibilities of improvement. Furthermore, we
aim to provide an open source implementation of the frameworks.
The master thesis project would include the following stages:
1. Collect the required bibliography and understand the basic theoretical context in each framework.
2. Implement at least one algorithm of each framework using the same programming language so that cross-experimentation makes
sense.
3. Design and propose experiments to get interesting comparison between the implementations.
4. Draw a conclusion out of the study.
References
[1] Saso Dzeroski. "Multi-relational data mining: an introduction". In: ACM SIGKDD Explorations Newsletter 5.1 (July 2003), pp.
1-16. issn: 1931-0145. doi: 10.1145/959242.959245. url: https://doi.org/10.1145/959242.959245 (visited on 04/04/2022).
Keywords:
Knowledge formalisation, Multi-relational data mining, RCA, PCA, G-FCA
Conditions:
The contract will be for six months, from 15/11/2022 to 15/05/2023
The employer will be the University of Lorraine
The salary is 600 euros per month
The expected profile is that of a student able to code with different programming languages and with a very solid foundation in
advanced mathematics.
The employer will be the University of Lorraine
The salary is 600 euros per month
The expected profile is that of a student able to code with different programming languages and with a very solid foundation in
advanced mathematics.
Department(s):
| Eco-Technic systems engineering |
Funds:
The funding comes from a CIFRE PhD project.
