Those knowledge representation forms can be considered traditional, because they are related with historically first times, mainly symbolic period of Artificial Intelligence. Are they applied so widely, to justify their inclusion in a general undergraduate course for Computer Science students?
Greetings professor and great question!
I do hope to see interesting answers here. I will contribute based on my brief experience in the area.
I think that we can see several classical knowledge representation forms in modern applications, for example semantic networks is really popular. In the other hand, vector representation seems the most widespread representation form for machine learning. Vectorial knowledge representation is a different (bottom-up) approach compared to semantic networks. Fussing the top-down approach of semantic networks with the conventional bottom-up approach of machine learning seems to be a current-to-future trend in the area.
No matter what, I think that is more important to discuss the limitations of vectorial data representation despite its widespread. Maybe that would be a good opportunity to explain how representation is one of the core issues in pattern recognition and the role of the sub discipline of learning representations from data (and the incredible hypo that deep learning have had in this area due to its ability to learn several layers of representation from raw sensorial data). The current trend is clear when enough data is available: skip manually handcrafted features and learn the representation automatically.
Hello,
in my opinion, you should always mention the roots of something you are going to use or explain to give a brief introduction where everything originated from. Sometimes things get more clear knowing the background of it. How much you focus on it depends on you or the application/your focus in the end. Sadly i cannot answer your question regarding the usage of such things.
Hope this helps a bit!
Regards,
André
Hi.,
A frame is a knowledge representation technique that used to represent knowledge using numbers of frames related to each other by relationship . It works based on the concept of inheritance. ... i.e An instance of frame is representation of a specific object. Every slot has name and corresponding value.
Thus, a KR-scheme (representational language) is a formal description of the symbols used for the representation, as well as their real world semantics. A KB is used to store the knowledge of the agent.
Greetings professor and great question!
I do hope to see interesting answers here. I will contribute based on my brief experience in the area.
I think that we can see several classical knowledge representation forms in modern applications, for example semantic networks is really popular. In the other hand, vector representation seems the most widespread representation form for machine learning. Vectorial knowledge representation is a different (bottom-up) approach compared to semantic networks. Fussing the top-down approach of semantic networks with the conventional bottom-up approach of machine learning seems to be a current-to-future trend in the area.
No matter what, I think that is more important to discuss the limitations of vectorial data representation despite its widespread. Maybe that would be a good opportunity to explain how representation is one of the core issues in pattern recognition and the role of the sub discipline of learning representations from data (and the incredible hypo that deep learning have had in this area due to its ability to learn several layers of representation from raw sensorial data). The current trend is clear when enough data is available: skip manually handcrafted features and learn the representation automatically.
Md Arafat Habib Thanks.
I think it depends on the perspective of the course. Usually people nowadays want easy go. But from my opinion, it can be a prerequisite for enrolling the AI courses.
Generally if you think , for solving the classification problem, usually one does not need to know the knowledge representation forms, like scripts, semantic networks and frames. But depth of knowledge is always needed.
Nevertheless their real-world applications it is worth to mention them (just few hours) as they are a part of symbolic AI. Then, it could be interesting to show how these historical methods can be combined with new methods coming from numeric AI, such as the machine learning algorithms.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Artificial Intelligence