Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
CogPro: Cognitive Processor for Astronomical Big Data Analysis Amit Kumar Mishra Electrical Engineering Department University of Cape Town, South Africa Email: [email protected] 1 Introduction Brain has intrigued researchers since the beginning of scientific endeavors. Firstly, beginning of computers saw the advent of exciting developments which culminated to the development of the new discipline of artificial neural networks (ANN). ANNs have been through several generations of major developments, with the recent phase consisting of spiking neural networks based works [1]. Another parallel field of computational neuroscience has been the bio-inspired cognitive architectures (BICA) [2] a field which got major thrust in development. Cognitive architecture (CA) in general and BICA in particular also has a long history and the efforts have been devoted towards trying to emulate the functioning of brain. CAs like SOAR and ACT-R have been under development for many decades and have been applied in various studies [3, 4]. A third direction in cognitive engineering has been the recent developments in communication and radar which are misleadingly termed cognitive communication [5] and cognitive radar [6]. It may also be mentioned here that this 2012-2013 has seen multi-billion dollar investment done separately in the European Union as well as in the USA for the study and understanding of brain [7–9]. 1.1 Brain Inspired Hardware One of the major conceptual background to BICA systems is their layered architecture with each higher layer representing higher abstraction information. There are many running and planned work in such kind of systems. In terms of computational platform and ASICs, following products are note-worthy. • Neurogrid – Mixed analog-digital electronics [10] • Neuromem – Neuron inspired digital electronics http : //www.electronics−eetimes.com/news/neuromem−ic−matches− patterns − sees − all − knows − all/page/0/2 • IBM SyNAPSE – Memristor technology [11] 1 • Qualcomm Snapdragon 820 – Neural Processing Unit [12] • SpiNNaker – Spiking Neural Network architecture [13] In terms of software and algorithm developments the following two are noteworthy. • Google-DeepMind: This is a small company acquired by Google and they work on the development of BICA algorithms for Google. Most of their work and codes are available for free in their website. An example of their work can be found at [14]. • Numenta: This is a company that is developing the infrastructure to build BICA algorithms. They are keeping it open-source and relatively flexible [15]. In this proposed MSc project the student will comeup with the optimal processing architecture to run a BICA of choice, and implement it using the available resources. It can be added here that one of the major bottleneck for BICA implementation was the complexity of handling heterogenous computing platforms. Some of the upcoming chips like the planned Xeon (which will have both CPU with major FPGA slices in the same chip) will ease this bottleneck and hence make this proposed MSc project very pertinent. 2 Existing Expertise at UCT We at Electrical Engineering Department at UCT have been active in areas related to cognitive radio and cognitive radar for past six years. For last two years the proposer has also been working on cognitive robotics. In addition we have some of the best South African experts in the domain of mathematical modelling [16] and psychological understanding of brain [17]. The recently opened Center of Artificial Intelligence is also centered at UCT (Computer Science Department). We also have close ties with the Center for High Performance Computing (CHPC), South Africa, which hosts some of the best high performance platforms in the continent. These and the existing rich link our group has with SKA-SA makes us the best place to host activities to design hardware platforms to host cognitive architectures. 3 Conclusion Bigdata analysis is a rich field with manifold of approaches in it. However BICA is one of the most revolutionary approaches that has been proposed to solve this problem. The proposed project along with another MSc project proposed by us will not only have very high probability to become a tool that will really be used by SKA, it can also be a major engineering spin-off of the SKA project. References [1] J. D. Bransford, A. L. Brown, and R. R. Cocking, How people learn: Brain, mind, experience, and school. National Academy Press, 1999. 2 [2] A. V. Samsonovich, “Toward a unified catalog of implemented cognitive architectures.” BICA, vol. 221, pp. 195–244, 2010. [3] P. S. Rosenbloom, J. E. Laird, A. Newell, and R. McCarl, “A preliminary analysis of the soar architecture as a basis for general intelligence,” Artificial Intelligence, vol. 47, no. 1, pp. 289–325, 1991. [4] N. A. Taatgen, C. Lebiere, and J. R. Anderson, “Modeling paradigms in act-r,” Cognition and multi-agent interaction: From cognitive modeling to social simulation, pp. 29–52, 2006. [5] J. Mitola, “Cognitive radio—an integrated agent architecture for software defined radio,” 2000. [6] S. Haykin, “Cognitive radar: a way of the future,” Signal Processing Magazine, IEEE, vol. 23, no. 1, pp. 30–40, 2006. [7] H. Markram, “The blue brain project,” Nature Reviews Neuroscience, vol. 7, no. 2, pp. 153–160, 2006. [8] ——, “The human brain project,” Scientific American, vol. 306, no. 6, pp. 50–55, 2012. [9] R. Knotts, “Overview of federal funding agency priorities and interdisciplinary themes,” 2013. [10] B. V. Benjamin, P. Gao, E. McQuinn, S. Choudhary, A. R. Chandrasekaran, J.-M. Bussat, R. Alvarez-Icaza, J. V. Arthur, P. A. Merolla, and K. Boahen, “Neurogrid: A mixed-analog-digital multichip system for large-scale neural simulations,” Proceedings of the IEEE, vol. 102, no. 5, pp. 699–716, 2014. [11] G. Indiveri, B. Linares-Barranco, R. Legenstein, G. Deligeorgis, and T. Prodromakis, “Integration of nanoscale memristor synapses in neuromorphic computing architectures,” Nanotechnology, vol. 24, no. 38, p. 384010, 2013. [12] F. J. Ordóñez and D. Roggen, “Deep convolutional and lstm recurrent neural networks for multimodal wearable activity recognition,” Sensors, vol. 16, no. 1, p. 115, 2016. [13] M. M. Khan, D. R. Lester, L. A. Plana, A. Rast, X. Jin, E. Painkras, and S. B. Furber, “Spinnaker: mapping neural networks onto a massively-parallel chip multiprocessor,” in Neural Networks, 2008. IJCNN 2008.(IEEE World Congress on Computational Intelligence). IEEE International Joint Conference on. Ieee, 2008, pp. 2849–2856. [14] A. Graves, G. Wayne, and I. Danihelka, “Neural turing machines,” arXiv preprint arXiv:1410.5401, 2014. [15] D. George, “How the brain might work: A hierarchical and temporal model for learning and recognition,” Ph.D. dissertation, Stanford University, 2008. 3 [16] N. Murphy, G. Ellis, and T. O’Connor, Downward causation and the neurobiology of free will. Springer Science & Business Media, 2009. [17] J. B. Savitz, M. Solms, and R. S. Ramesar, “Neurocognitive function as an endophenotype for genetic studies of bipolar affective disorder,” Neuromolecular Medicine, vol. 7, no. 4, pp. 275–286, 2005. 4