1C. Computational Neuroscience, Machine Learning

---1--- xxx

---2--- xxx

---3--- xxx

---4--- xxx

--3--- Computational Neuroscience & Deep Machine Learning

‘Computational Neuroscience’ and ‘Deep Machine Learning’ (CN/DML) are very exciting interdisciplinary fields of science. These fields of study involve using physics, mathematics, engineering and computational science, to understand how biological nervous systems work, in terms of their structures and functions, at every level of structural organisation (from the molecular/cellular, to networks of neural cells, to whole organ systems). However, these fields of study can also refer to how we can use our understanding of neurobiological systems to give computers (made of silicon-semiconductor-based transistors) the ability to operate more like human brains, especially when it comes to ‘learning’. Mathematical statisticians, computer engineers and software programmers may refer to these fields of study more broadly as ‘data science’, while biological scientists and bioanalysts may refer to them as ‘neuro-cognitive physiology’. Most people may simply refer to these cognitive functions of machines, as ‘General-purpose Artificial Intelligence’ (GAI/AGI). The more science-savvy readers of this will of course debate over the definition of all these terms. However, this debate can be put aside for now, as there is certainly a lot of overlap between all these terms.

AGI programs are mainly based on Artificial Neural Networks (ANNs). ANNs are computational and mathematical models inspired by biological neural networks; and they also have a physical implementation. Connectivity and interactions of the neural nodes is important. Nodes (standard artificial neurons) can be organised and connected in such a way, as to encode & store & process data, in particular ways. Certain ANNs represent algorithms that have the ability to learn and perform certain computations:

● The algorithms are not pre-programmed to solve a particular task, they learn how to solve a particular task, just like the human brain.

● The learning is based on forming associations between node A and node B. This is done by increasing (potentiating) or decreasing (depressing) the connection weights (synaptic strengths) between neural nodes.

● Furthermore, the algorithms are not narrowly-defined or domain-specific, they are general-purpose. The same algorithm will perform well across a range of different tasks.

● So in other words, all the rules don’t have to be programmed in manually by the programmer; the ANNs learn to abstract their own rules, and self-modify their own high-level code using learning rules.

● Given enough processing power (by using more layers and nodes; aka. ‘deeper’ neural networks), they do considerably better when it comes to optimization and processing ‘big-data’, compared to conventional methods. They are a massive extension of the algorithms we have had up until the mid-2000s, which were just a combination of human-made computer-based non-learning automata and expert human cognition, but not true AGI.

● ANNs can provide more optimised optimal solutions, and to a wide range of complex problems.

● There is supervised learning and unsupervised learning, and everything in between.

***

●●● click: eLearning video resources by HM ●●●

***

Figures/Diagrams (5) Below are by H Muzart

---

Here are some strongly recommended links to other sources:

● www.bioneurotech.com > products-services

● https://www.sciarticles.wordpress.com/article-2/

● http://www.gatsby.ucl.ac.uk/research.html (UCL Computational Neuroscience Unit)

● www.icn.ucl.ac.uk (Institute of Cognitive Neuroscience - UCL)

● www.ucl.ac.uk/swc (Sainsbury Wellcome Centre for Neural Circuits and Behaviour - UCL)

● http://cs229.stanford.edu/materials.html (Machine Learning info)

● https://ml.berkeley.edu/ (UC Berkeley Machine Learning)

● machinelearning.mit.edu/

● www.seas.harvard.edu/courses/cs281/

● https://www.coursera.org/learn/computational-neuroscience

● https://www.khanacademy.org/science/health-and-medicine/nervous-system-and-sensory-infor (Advanced nervous system physiology)

● https://www.imperial.nhs.uk/our-services/neurology (Imperial College Healthcare | Neurology)

● https://deepmind.com/research/publications/ (AI company, OA research papers)

● www.kcl.ac.uk/ioppn/depts/bcn/ (King's College London - Department of Basic & Clinical Neuroscience)

● www.brainnet.net/about/brain-resource-international-database/ (The BRAINnet Database - Brain Research And Integrative Science)

● https://crcns.org/ (CRCNS database data-sharing website)

● https://openeuroscience.com/data-collaboration.../neuroimaging-database (Neuroimaging databases | Open neuroscience)

● www.jneurosci.org/content/22/5/1497.long (Databases for Neuroscience Research)

● forum2016.fens.org/ (10th FENS Forum of Neuroscience 2016)

● neurosciencenews.com/

● groupspaces.com/ICNSubjectDatabase/ (ICNSubjectDatabase - GroupSpaces Expts & Clinical trials)

● https://www.ucl.ac.uk/complex (Centre for Mathematics and Physics in the Life Sciences and EXperimental Biology)

● neurosci.umin.jp/e/ (The University of Tokyo - Graduate School of Integrative Medical Neuroscience)

● https://www.ethz.ch/.../master-neural-systems-and-computation.html (Neural Systems and Computation | ETH Zurich)

● http://www.cs.ucl.ac.uk/teaching_learning/ (Computer engineering UCL - resources)

● http://theopendata.com/site/

● https://doaj.org/ (Directory of Open Access Journals)

● https://polymathprojects.org/ (OA collaborative mathematical & scientific projects)

● brain-map.org/ (OA database of gene expression patterns, by Allen Brain Institute)

● https://openai.com/about/ (OpenAI work, involvement by Elon Musk)

● www.GitHub.com (crowd-sourcing public open source source-code)

● https://cloud.google.com/ml/ (Google Developers labs; OA Cloud Machine Learning platform using Tensorflow framework - for predictive analytics)

● http://www.complexity.ecs.soton.ac.uk ; www2.warwick.ac.uk/fac/cross_fac/complexity (complexity science & data systems)

● https://leonardoaraujosantos.gitbooks.io/artificial-inteligence/content/machine_learning.html

###########

If you want to find out more about companies using computational neuroscience / deep machine learning:

AI Companies listings (names, net worth, locations, etc) (Websites)

● http://www.ai-one.com (list of AI companies)

● http://www.datamation.com/applications/top-20-artificial-intelligence-companies.html

● https://index.co/market/artificial-intelligence/companies

● https://angel.co/artificial-intelligence (list of AI companies)

● https://www.theguardian.com/technology/2014/jan/27/google-acquires-uk-artificial-intelligence-startup-deepmind

● https://www.cbinsights.com/blog/top-acquirers-ai-startups-ma-timeline/

● https://www.tractica.com/newsroom/press-releases/artificial-intelligence-technologies-are-quietly-penetrating-a-wide-range-of-enterprise-applications/

● https://www.wired.com/2012/10/open-science-roller-coaster-accelerates/

● https://techcrunch.com/2016/01/21/the-world-economic-forum-on-the-future-of-jobs/

[BackToTop]

Here are some videos by other people; these videos have been selected as they are excellent when it comes to getting some understanding of basic concepts in this topic:

"What is Computational Neuroscience?" by BernsteinCenterFR

[Aug 2011] [4min10sec]

http://youtube.com/watch?v=d5oqIxhTU8I