6 Neural-symbolic learning and reasoning – Besold et al. (2017)
Dit paper bespreekt neural symbolic modellen
• De auteurs maken onderscheidt tussen connectionistische representaties en symbolische
representaties. Wat zijn deze en hoe verschillen ze van elkaar?
• Neural symbolic modellen proberen connectionistische en symbolische representaties
samen te voegen in 1 model. Hoe doen ze dit? Welke rol spelen beiden?
- behavior is determined and guided by cognition and mental processing, pominent tools in the
modelling of behavior are: computational-logic systems, connectionist models of cognition and
the brain, models of uncertainty
- in neural computing, it is assumed that the mind is an emergent property of the brain, and that
computational cognitive modelling can lead to valid theories of cognition and offer an
understanding of certain cognitive processes
→ connectionism should be able to offer an appropriate representational language for AI as well
→ connectionism: an artificial intelligence approach to cognition in which multiple connections
between nodes (equivalent to brain cells) form a massive interactive network in which many
processes take place simultaneously and certain processes, operating in parallel, are grouped
together in hierarchies that bring about results such as thought or action.
- logic is seen as a fundamental tool in the modelling of thought and behaviour
- the methodology of neural-symbolic systems aims to transfer principles and mechanisms
between logic-based computation and neural computation
→ it considers how principles of symbolic computation can be implemented by connectionist
mechanisms and how subsymbolic computation can be described and analysed in logical terms
→ connectionism provides the hardware upon which different levels of abstraction can be built
according to the needs of the application
- subsymbolic computation refers to processing information at a level lower than symbolic
representations. its more about working with implicit, non-symbolic representations. It can be
applied in machine learning.
- neural-symbolic system: artificial neural networks (ANN) provide the machinery for parallel
computation and robust learning, while logic provides the necessary explanation for the network
models, facilitating the necessary interaction with the world and other systems
- rational agents: perform concept acquisition (generally unsupervised and statistical) and
concept manipulation (generally supervised and symbolic) as part of a permanent cycle of
perception and action
- neural-symbolic integration is seen as a way of addressing the challenge through the
mechanisms of knowledge translation and knowledge extraction between symbolic logic
systems and subsymbolic networks
- the merching of theory and data learning in ANNs is more effective than purely symbolic or
purely connectionist systems (especially in the case of real-world, unstructured data)
Dit paper bespreekt neural symbolic modellen
• De auteurs maken onderscheidt tussen connectionistische representaties en symbolische
representaties. Wat zijn deze en hoe verschillen ze van elkaar?
• Neural symbolic modellen proberen connectionistische en symbolische representaties
samen te voegen in 1 model. Hoe doen ze dit? Welke rol spelen beiden?
- behavior is determined and guided by cognition and mental processing, pominent tools in the
modelling of behavior are: computational-logic systems, connectionist models of cognition and
the brain, models of uncertainty
- in neural computing, it is assumed that the mind is an emergent property of the brain, and that
computational cognitive modelling can lead to valid theories of cognition and offer an
understanding of certain cognitive processes
→ connectionism should be able to offer an appropriate representational language for AI as well
→ connectionism: an artificial intelligence approach to cognition in which multiple connections
between nodes (equivalent to brain cells) form a massive interactive network in which many
processes take place simultaneously and certain processes, operating in parallel, are grouped
together in hierarchies that bring about results such as thought or action.
- logic is seen as a fundamental tool in the modelling of thought and behaviour
- the methodology of neural-symbolic systems aims to transfer principles and mechanisms
between logic-based computation and neural computation
→ it considers how principles of symbolic computation can be implemented by connectionist
mechanisms and how subsymbolic computation can be described and analysed in logical terms
→ connectionism provides the hardware upon which different levels of abstraction can be built
according to the needs of the application
- subsymbolic computation refers to processing information at a level lower than symbolic
representations. its more about working with implicit, non-symbolic representations. It can be
applied in machine learning.
- neural-symbolic system: artificial neural networks (ANN) provide the machinery for parallel
computation and robust learning, while logic provides the necessary explanation for the network
models, facilitating the necessary interaction with the world and other systems
- rational agents: perform concept acquisition (generally unsupervised and statistical) and
concept manipulation (generally supervised and symbolic) as part of a permanent cycle of
perception and action
- neural-symbolic integration is seen as a way of addressing the challenge through the
mechanisms of knowledge translation and knowledge extraction between symbolic logic
systems and subsymbolic networks
- the merching of theory and data learning in ANNs is more effective than purely symbolic or
purely connectionist systems (especially in the case of real-world, unstructured data)
