Predictive Processing
Overall predictive processing is a large and varied field of study which takes the perspective that prediction is core to an organism’s perceptual system.
We take a predictive perspective in Reinforcement Learning, as all behavior is driven by maximizing the expected sum of future Reward. While this drives most of the development in Representation learning indirectly through gradient descent on the Value Function or a set of Auxiliary Tasks. This is typically discussed in terms of the Predictive Knowledge framework (White 2015; Kearney and Oxton 2019)
Other takes on predictive processing can be found by accounts:
- (Clark 2013): This is an extension of the ideas presented in (Rao and Ballard 1999), thinking of the consequences of such a framework.
- PSR’s (i.e. predictive representations of state) which include a wide range of ideas.
Projects
Questions
Literature
References
Clark, Andy. 2013. “Whatever next? Predictive Brains, Situated Agents, and the Future of Cognitive Science.” Behavioral and Brain Sciences.
Kearney, Alex, and Oliver Oxton. 2019. “Making Meaning: Semiotics Within Predictive Knowledge Architectures.” arXiv. https://arxiv.org/abs/1904.09023.
Links to this note:
- Developmental Reinforcement Learning
- bubic2010prediction: Prediction, cognition and the brain
- clark2013whatever: Whatever next? Predictive brains, situated agents, and the future of cognitive science
- huang2011predictive: Predictive Coding
- synofzik2013experience: The experience of agency: an interplay between prediction and postdiction
- stock2004short: A short history of ideo-motor action
- soga2009predictive: Predictive and postdictive mechanisms jointly contribute to visual awareness
- rao1999predictive: Predictive coding in the visual cortex: a functional interpretation of some extra-classical receptive-field effects
- Free-Energy Principle