CEN launch, Gert Westermann
Download as PPT, PDF1 like394 views
Neil Burgess' talk discussed using computational modelling in neuroscience and education. Models can help characterize how learning changes the brain by developing a model that displays the same behaviour as observed in the brain or behaviourally. This provides a theory for how the process works. Models can also explore constraints on learning and change from factors like environment, genes, and the developing brain. Models should be constrained by neuroscience findings on mechanisms of synaptic change and how brain regions interact to produce behaviour. Finding an appropriate level of abstraction in models may bridge the gap between neuroscience and psychology.
CEN launch, Gert Westermann
- 1. Neuroscience, Computational Modelling and Education: Reflections on Neil Burgess’ talk Gert Westermann
- 2. Modelling already featured in Neil’s talk: x j x i w ij e.g. w ij -> w ij + ε x j x i w ij -> w ij + Δ w ij x j w j Different learning rules in hippocampus and striatum?
- 3. So what can modelling offer to neuroscience and education?
- 4. BEHAVIOUR Huge gap! Computational models
- 5. Models… … help us to understand how learning changes the brain (to characterize the process of change) Basic idea: We observe a process (e.g., brain-behaviour correlates), or more relevant here, a behavioural change We develop a computational model that displays the ‘same’ behaviour We know how the model works, and this becomes our theory of how the process works in real life But this is not always followed. /rIt/ write
- 6. Neural network (connectionist) models Added (important) benefit: Functionality of these models is inspired by how neurons work Although we should stay alert to the limits of this analogy.
- 7. Characterizing constraints on change Models as a tool to explore what affects change: Environment frequency of exposure order of exposure (age of acquisition) type of exposure (e.g., similarity between stimuli)
- 8. Characterizing constraints on change Genes/internal constraints Structure/resources of the learning system (critical periods, developmental disorders, speed-accuracy trade-off in learning) ε = 0.1
- 9. Characterizing constraints on change Links between brain and cognitive development Effect of environmental exposure on development of functional structures Effect of the integration of subsystems on behaviour Maturation and experience-dependent plasticity
- 10. These aspects of models should be constrained by neuroscience: Mechanisms of synaptic change Interplay of functional brain regions and give rise to relevant behaviour. Converging evidence
- 11. Bridging the gap… Models can be built at different levels of abstraction. Is there a level that is acceptable both to neuroscientists and psychologists? I think: yes, if we constantly remind ourselves what a model is for.