GEMM-SaFIN(FRIE)++ : explainable artificial intelligent with episodic memory
Neuro-fuzzy systems are hybrid systems that take advantage on the functionalities of fuzzy logics and neural networks. The IF-THEN fuzzy rules allow good interpretability for human experts to understand the correlation between inputs and outputs. However, only the neuro-fuzzy’s designer knows the me...
| Prif Awdur: | |
|---|---|
| Awduron Eraill: | |
| Fformat: | Final Year Project (FYP) |
| Iaith: | English |
| Cyhoeddwyd: |
Nanyang Technological University
2020
|
| Pynciau: | |
| Mynediad Ar-lein: | https://hdl.handle.net/10356/137999 |
| _version_ | 1826111067190722560 |
|---|---|
| author | Ko, Nelson Mingwei |
| author2 | Quek Hiok Chai |
| author_facet | Quek Hiok Chai Ko, Nelson Mingwei |
| author_sort | Ko, Nelson Mingwei |
| collection | NTU |
| description | Neuro-fuzzy systems are hybrid systems that take advantage on the functionalities of fuzzy logics and neural networks. The IF-THEN fuzzy rules allow good interpretability for human experts to understand the correlation between inputs and outputs. However, only the neuro-fuzzy’s designer knows the mechanism and behavior of the system. To an enterprise point of view, e.g. finance world, the details on what makes the system arrive and formulate its predictions or rules is unknown. This is due to the lack of transparency between the system and the human experts.
This paper proposes two novel types of explainable artificial intelligent (XAI) feature implemented using the neuro fuzzy architecture called Self Adaptive Fuzzy Inference Network with Fuzzy Rule Interpolation or Extrapolation (SaFIN(FRIE)). The two explainable AI feature are in a form of a graphical user interface (GUI) to assist the human expert to understand the inner mechanics of function on how it draws conclusion with the data fed into the system.
One of the challenges for the fuzzy neural system is the making of a real-time prediction in the financial market where data can be sparse. As such, it is not able to automatically detect and react to the occurrence of concept drift and shifts, affecting their online learning capabilities. SaFIN(FRIE) employs interpolation and extrapolation techniques to make inference when drift or shift is detected. A general episodic memory technique is employed to capture and retrieve from past events that SaFIN(FRIE) learns. This is done by storing and retrieving them from an episodic memory storage during transient event behavior. |
| first_indexed | 2024-10-01T02:44:32Z |
| format | Final Year Project (FYP) |
| id | ntu-10356/137999 |
| institution | Nanyang Technological University |
| language | English |
| last_indexed | 2024-10-01T02:44:32Z |
| publishDate | 2020 |
| publisher | Nanyang Technological University |
| record_format | dspace |
| spelling | ntu-10356/1379992020-04-21T09:03:55Z GEMM-SaFIN(FRIE)++ : explainable artificial intelligent with episodic memory Ko, Nelson Mingwei Quek Hiok Chai School of Computer Science and Engineering ashcquek@ntu.edu.sg Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence Engineering::Computer science and engineering::Software::Software engineering Neuro-fuzzy systems are hybrid systems that take advantage on the functionalities of fuzzy logics and neural networks. The IF-THEN fuzzy rules allow good interpretability for human experts to understand the correlation between inputs and outputs. However, only the neuro-fuzzy’s designer knows the mechanism and behavior of the system. To an enterprise point of view, e.g. finance world, the details on what makes the system arrive and formulate its predictions or rules is unknown. This is due to the lack of transparency between the system and the human experts. This paper proposes two novel types of explainable artificial intelligent (XAI) feature implemented using the neuro fuzzy architecture called Self Adaptive Fuzzy Inference Network with Fuzzy Rule Interpolation or Extrapolation (SaFIN(FRIE)). The two explainable AI feature are in a form of a graphical user interface (GUI) to assist the human expert to understand the inner mechanics of function on how it draws conclusion with the data fed into the system. One of the challenges for the fuzzy neural system is the making of a real-time prediction in the financial market where data can be sparse. As such, it is not able to automatically detect and react to the occurrence of concept drift and shifts, affecting their online learning capabilities. SaFIN(FRIE) employs interpolation and extrapolation techniques to make inference when drift or shift is detected. A general episodic memory technique is employed to capture and retrieve from past events that SaFIN(FRIE) learns. This is done by storing and retrieving them from an episodic memory storage during transient event behavior. Bachelor of Engineering (Computer Science) 2020-04-21T09:03:54Z 2020-04-21T09:03:54Z 2020 Final Year Project (FYP) https://hdl.handle.net/10356/137999 en SCSE19-0530 application/pdf Nanyang Technological University |
| spellingShingle | Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence Engineering::Computer science and engineering::Software::Software engineering Ko, Nelson Mingwei GEMM-SaFIN(FRIE)++ : explainable artificial intelligent with episodic memory |
| title | GEMM-SaFIN(FRIE)++ : explainable artificial intelligent with episodic memory |
| title_full | GEMM-SaFIN(FRIE)++ : explainable artificial intelligent with episodic memory |
| title_fullStr | GEMM-SaFIN(FRIE)++ : explainable artificial intelligent with episodic memory |
| title_full_unstemmed | GEMM-SaFIN(FRIE)++ : explainable artificial intelligent with episodic memory |
| title_short | GEMM-SaFIN(FRIE)++ : explainable artificial intelligent with episodic memory |
| title_sort | gemm safin frie explainable artificial intelligent with episodic memory |
| topic | Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence Engineering::Computer science and engineering::Software::Software engineering |
| url | https://hdl.handle.net/10356/137999 |
| work_keys_str_mv | AT konelsonmingwei gemmsafinfrieexplainableartificialintelligentwithepisodicmemory |