![THE ART OF MAPPING DATA AND EQUATIONS
In 1968 [1], neural recordings from cat visual cortex demonstrated the encoding
of data to predict neural responses given a stimulus.
rmax: maximum average response rate (spikes/second, Hz)
f: firing rate (Hz)
s: stimulus orientation; σf: tuning curve width](https://image.slidesharecdn.com/cbccompneuroaimsmiia-171206153623/85/Chris-Currin-computational-neuroscience-intro-AIMS-MIIA-2017-12-5-320.jpg)
![NEUROSCIENCE AND AI
“only by better understanding human intelligence can we hope to push the
boundaries of what artificial intellects can achieve.” – Demis Hassabis
...understanding biological brains could
play a vital role in building intelligent
machines... Distilling intelligence into an
algorithmic construct and comparing it
to the human brain might yield insights
into some of the deepest and the most
enduring mysteries of the mind... [2]
[3]](https://image.slidesharecdn.com/cbccompneuroaimsmiia-171206153623/85/Chris-Currin-computational-neuroscience-intro-AIMS-MIIA-2017-12-15-320.jpg)
![TOP-DOWN UNDERSTANDING
• image computable
• mappable onto the visual
pathway, and
• predictive of neural
responses
[4]](https://image.slidesharecdn.com/cbccompneuroaimsmiia-171206153623/85/Chris-Currin-computational-neuroscience-intro-AIMS-MIIA-2017-12-21-320.jpg)
![IBRO-SIMONS COMPUTATIONAL NEUROSCIENCE IMBIZO
http://isicni.gatsby.ucl.ac.uk/
THANK YOU1. COMPUTATIONAL NEUROSCIENCE PROVIDES A FRAMEWORK FOR UNDERSTANDING THE BRAIN
2. WE SHOULD CARE BECAUSE IT CAN ADVANCE AND BETTER HUMANITY
3. WE CAN UNDERSTAND THE BRAIN USING TOP-DOWN AND BOTTOM-UP APPROACHES
[1]: Hubel, D.H. & Weisel, T.N. (1968) Receptive fields and functional architecture of themonkey striate cortex. Journal of Physiology 195:215-243
[2]: Hassabis, D., Kumaran, D., Summerfield, C., Botvinick M. (2017) Neuroscience-inspired artificial intelligence. Neuron 95: 245-258
[3]: Costa, R.P., Assael, Y.M., Shillingford, B., de Freitas, N., Vogels, T.P. (2017) Cortical microcircuits as gated-recurrent neural networks. arXiv:1711.02448
[4]: Yamins, D.L.K., and DiCarlo, J.J. (2016) Using goal-driven deep learning models to understand sensory cortex. Nature Neurscience. 19: 356-365
chris.currin@gmail.com](https://image.slidesharecdn.com/cbccompneuroaimsmiia-171206153623/85/Chris-Currin-computational-neuroscience-intro-AIMS-MIIA-2017-12-24-320.jpg)
Chris Currin computational neuroscience intro AIMS MIIA 2017-12
- 1. COMPUTATIONAL NEUROSCIENCE THINKING ABOUT THINKING • BSc Biochemistry, Psychology, Computer Science (Rhodes + UCT) • Visiting academic researcher 2016 (Tim Vogels group, CNCB, Oxford University) • PhD Computational Neuroscience candidate (Raimondo Lab, Human Biology, UCT) • “Intelligence and Intelligent Systems Specialist” – iXperience, Udacity, Q Division, NumberBoost
- 2. OUTLINE 1. WHAT IS COMPUTATIONAL NEUROSCIENCE? 2. WHY SHOULD WE CARE? 3. HOW CAN WE UNDERSTAND THE BRAIN?
- 3. NPC Brain Credit NICHD I. Williams DISCLAIMER: THIS IS BIOLOGY
- 4. DEFINITION Computational neuroscience research involves building mathematical and computer-based models that provide a theoretical framework that encapsulates our emerging understanding of the brain’s functions.
- 5. THE ART OF MAPPING DATA AND EQUATIONS In 1968 [1], neural recordings from cat visual cortex demonstrated the encoding of data to predict neural responses given a stimulus. rmax: maximum average response rate (spikes/second, Hz) f: firing rate (Hz) s: stimulus orientation; σf: tuning curve width
- 6. WHAT DOES THE BRAIN DO? Processing ~106 neurons ~106 neurons ~1011 neurons ?
- 7. HOW DOES IT WORK? sensory processing (input) motor control (output) action selection memory
- 8. HOW DOES THE BRAIN COMPUTE?
- 9. “WHY SHOULD I CARE?” Ok, but
- 11. • I bet you 11456324532 dollars you didn't read that number. • You just skipped right over it. • You didn't even realize I put a letter in it. • No I didn't but you went back and looked.
- 12. WHY SHOULD I CARE? • Framing and guiding experimental research Reduce use of animals in research • Help understand + cure diseases and disabilities • Advance AI and Data Science The brain processes information to give us a prediction of the world It’s not always right
- 13. WHAT DOES THE BRAIN COMPUTE?
- 14. THE COMPUTATIONS ARE NOISY Trials Time (10s) Neuron 2 Neuron 1 Institute of Cognitive Science, http://neuroxidence.com
- 15. NEUROSCIENCE AND AI “only by better understanding human intelligence can we hope to push the boundaries of what artificial intellects can achieve.” – Demis Hassabis ...understanding biological brains could play a vital role in building intelligent machines... Distilling intelligence into an algorithmic construct and comparing it to the human brain might yield insights into some of the deepest and the most enduring mysteries of the mind... [2] [3]
- 16. NEUROSCIENCE-INSPIRED AI Future • Intuitive understanding of the physical world • Transfer learning • zero-shot inferences (Higgins et al. 2016) • progressive network • Imagination and Planning • Virtual Brain Analytics Present • Attention • Episodic Memory • Working memory • Continual learning http://www.cell.com/neuron/fulltext/S0896-6273(17)30509-3
- 17. HOW CAN WE UNDERSTAND THE BRAIN?
- 18. 19 • The brain contains lots and lots of neurons (1011). • And even more connections (1014). • The neurons communicate via spikes.
- 19. HOW CAN WE UNDERSTAND INTELLIGENCE? computational algorithmic implementation development evolution Marr’s ‘classical’ levels of analysis problem application solution learning meta
- 21. TOP-DOWN UNDERSTANDING • image computable • mappable onto the visual pathway, and • predictive of neural responses [4]
- 22. HOW CAN WE UNDERSTAND INTELLIGENCE? computational algorithmic implementation development evolution Marr’s ‘classical’ levels of analysis problem application solution learning meta
- 24. IBRO-SIMONS COMPUTATIONAL NEUROSCIENCE IMBIZO http://isicni.gatsby.ucl.ac.uk/ THANK YOU1. COMPUTATIONAL NEUROSCIENCE PROVIDES A FRAMEWORK FOR UNDERSTANDING THE BRAIN 2. WE SHOULD CARE BECAUSE IT CAN ADVANCE AND BETTER HUMANITY 3. WE CAN UNDERSTAND THE BRAIN USING TOP-DOWN AND BOTTOM-UP APPROACHES [1]: Hubel, D.H. & Weisel, T.N. (1968) Receptive fields and functional architecture of themonkey striate cortex. Journal of Physiology 195:215-243 [2]: Hassabis, D., Kumaran, D., Summerfield, C., Botvinick M. (2017) Neuroscience-inspired artificial intelligence. Neuron 95: 245-258 [3]: Costa, R.P., Assael, Y.M., Shillingford, B., de Freitas, N., Vogels, T.P. (2017) Cortical microcircuits as gated-recurrent neural networks. arXiv:1711.02448 [4]: Yamins, D.L.K., and DiCarlo, J.J. (2016) Using goal-driven deep learning models to understand sensory cortex. Nature Neurscience. 19: 356-365 chris.currin@gmail.com