BEGIN:VCALENDAR VERSION:2.0 PRODID:-//Memento EPFL// BEGIN:VEVENT SUMMARY:Neuromorphic Systems Research at IBM DTSTART:20150610T140000 DTEND:20150610T170000 DTSTAMP:20260410T232802Z UID:5444eedf129b2b3320812b27358dd832d5e55ee666525719995acebd CATEGORIES:Conferences - Seminars DESCRIPTION:Chung Lam\, Kamil Rocki\, Geoffrey W. Burr from IBM Research\n MINI-SYMPOSIUM JOINTLY HOSTED BY THE INSTITUTE OF ELECTRICAL ENGINEERING A ND THE INSTITUTE OF BIOENGINEERING\nThree top researchers from IBM will ta lk about the most recent advancements in this field. Each presentation wil l be followed by a short discussion session. The schedule is given below. Please see attached pdf for the complete schedule and speaker biographies. Schedule of the special event :\n14:00-14:40  Neuromorphic Engineering by Chung Lam\, IBM Research\, Yorktown Heights\, NY\, USAAbstract: Microproc essors designed with von Neumann architecture are hitting the power and pe rformance limits as silicon CMOS continues to scale the critical         dimensions of the circuit components towards single digit nanometer s ize limit. Multi-core processor\, parallel processing without increasing o perating frequency of the cores\, was introduced in the early 2000 to exte nd the power and performance scaling\, keeping Moore’s Law viable. Amdah l’s Law\, however\, argues that the performance speedup with parallel pr ocessing is governed by the percentage of algorithm that needed be serial. Evolution has provided us with the most efficient parallel processing arc hitecture: the biological brain. In this talk\, we shall examine what we c an do with little that we know about how the brain works to design machine s to mimick the brain.\n14:40-14:50  Questions/Discussion\n14:50-15:00  Coffee break\n15:00-15:40  HTM-based Saccadic Vision System by Kamil Rock i\, IBM Research–Almaden\, San Jose\, CA\, USAAbstract: In this project\ , Hierarchical Temporal Memory is used for rapid object categorization and tracking. Various studies have demonstrated the remarkable speed and effi ciency with which humans process natural scenes. Despite the fact that our eyes' fovea region is very limited\, we can efficiently view the world by redirecting the fovea between points of interest using eye movements call ed saccades. Using HTM\, we are able to learn both simple spatial patterns representing such small fovea region\, as well as predictable and invaria nt temporal patterns comprising whole sequence of saccades. Such an approa ch has two advantages: first\, storing images as temporal sequences of sma ll spatial building blocks is much more resource efficient than storing en tire complex images. The second one is that there is no need to distinguis h between storing and recognizing still and moving images.\n15:40-15:50  Questions/Discussion\n15:50-16:00  Coffee break\n16:00-16:40  Crossbar A rrays for Storage Class Memory and non-Von Neumann Computing by Geoffrey W . Burr\, IBM Research–Almaden\, San Jose\, CA\, USA\nFor more than 50 ye ars\, the capabilities of Von Neumann-style information processing systems - in which a "memory" delivers operations and then operands to a dedicate d "central processing unit" - have improved dramatically.  While it may s eem that this remarkable history was driven by ever-increasing density (Mo ore's Law)\, the actual driver was Dennard's Law: a device-scaling methodo logy which allowed each generation of smaller transistors to actually perf orm better\, in every way\, than the previous generation. Unfortunately\, Dennard's Law terminated some years ago\, and as a result\, Moore's Law is now slowing considerably. In a search for ways to continue to improve com puting systems\, the attention of the IT industry has turned to Non-Von Ne umann algorithms\, and in particular\, to computing architectures motivate d by the human brain.\nAt the same time\, memory technology has been going through a period of rapid change\, as new nonvolatile memories (NVM) - su ch as Phase Change Memory (PCM)\, Resistance RAM (RRAM)\, and Spin-Torque- Transfer Magnetic RAM (STT-MRAM) - emerge that complement and augment the traditional triad of SRAM\, DRAM\, and Flash.  Such memories could enable Storage-Class Memory (SCM) - an emerging memory category that seeks to co mbine the high performance and robustness of solid-state memory with the l ong-term retention and low cost of conventional hard-disk magnetic storage .\nSuch large arrays of NVM can also be used in non-Von Neumann neuromorph ic computational schemes\, with device conductance serving as the plastic (modifiable) “weight” of each “native” synaptic device.  This is an attractive application for these devices\, because while many synaptic weights are required\, requirements on yield and variability can be more r elaxed. However\, work in this field has remained highly qualitative in na ture\, and slow to scale in size.\nI will discuss our recent work towards large crossbar arrays of NVM for both of these applications. After briefly reviewing earlier work on PCM\, SCM\, and access devices based on copper- containing Mixed-Ionic-Electronic-Conduction (MIEC)\, I will discuss our r ecent work on quantitatively assessing the engineering tradeoffs inherent in NVM-based neuromorphic systems.\n16:40-16:50  Questions/Discussion LOCATION:SV1717a http://map.epfl.ch/?room=sv1717a STATUS:CONFIRMED END:VEVENT END:VCALENDAR