Emerging Technology: Cognitive Computing with Emerging Nanodevices

Members: Bilel Belhadj, Luis Camunas, Christian Denk, Doo Seok Jeong, Daniel Neil, Kang Wang, Hector Jesus Cabrera Villaseor, Jens Burger, James OSullivan, Jonathan Tapson, Mehmet Ozdas, Amir Khosrowshahi, Omid Kavehei, Sergio Davies, Patrick Sheridan, Shih-Chii Liu, Evangelos Stromatias, Shimeng Yu, Tara Julia Hamilton, Themistoklis Prodromakis, Timmer Horiuchi, Tobi Delbruck, Will Constable, Yezhou Yang

Leader: Omid Kavehei (University of Melbourne and Bionic Vision Australia) Tara Julia Hamilton (University of Western Sydney)

Invitees: Kang Wang (University of California, Los Angeles), Themistoklis Prodromakis (University of Southampton) and Doo Seok Jeong (Korea Institute of Science and Technology)

Subscribe to this special invited tutorial workgroup to learn about emerging nonvolatile resistive memories and their application in cognitive computing. Discussions will be around experiments on resistive memory (RRAM) technologies -including Redox-based Resistive Memory (ReRAM) and Conductive Bridge Memory (CBRAM)-, Phase Change Memory (PCM), Spin-Transfer Torque Magnetic Memory (STT-MRAM) and Magnetoelectric Memory (MeRAM).

This group will focus on practical techniques that allow successful implementation of artificial learning rules on nanocrossbar arrays. Highlighting feasibility and benefits of such approaches are the key messages of this group. The group considers pros and cons of designing nano-neuromorphic circuits and systems and provides clearer picture of different abstract levels (from materials and underlying physics to system level) for cognitive computing applications.

The central goal of this topic area is to gather neuromorphic circuit designers and participants who are keen to learn about emerging nonvolatile memory technologies. We aim to discuss previous implementations, the pros and cons of various techniques and ways in which algorithms and circuits from neuroscience can be implemented more efficiently and crossbar-friendly. The group will also encompass discussion groups on future directions and challenges in nano-neuromorphic design. We also aim to encourage further collaboration beyond Telluride.

A list of potential specific topic areas:

Energy scaling challenge of CMOS, problems with standby leakage and emerging nanodevice relative advantages in terms of nonvolatility, low switching energy, high speed, high endurance and scalability

Recent advances and scaling limits of the emerging memory technologies as well as new development to reduce the programming current and energy

Selection device challenge and architectural limitations

Cognitive computing with plastic solid state synapses

Device models for analog and digital applications

Stochastic learning with binary synapses using extremely low-power (sub-fJ) digital switches

Computing with fixed weights

Demonstration of complex synaptic behavior with a single RRAM device

Making reliable systems with unreliable devices

Talks and tutorials include: (some of the presentation files are attached)

Nonvolatile Nanoelectronics: From STT to MeRAM, Nonvolatile Intelligent Systems (by Kang Wang)

In order to scale down the switching energy, a new concept of electric-field control of magnetism, or voltage-controlled magnetoelectric (ME) memory (Me-RAM) is developed. Only recently, it is shown to be possible to use electric field to control metallic magnetism. For the latter, we will describe a couple of fundamental mechanisms of voltage control of magnetic moment and direction. With further advances in this area, we anticipate that the energy reduction of several orders of magnitudes beyond that of STT. Dynamics of the switching as well as additional physical processes in improving the switching process will be discussed.

Challenges and Opportunities of ReRAM for Neuromorphic Engineering (by Themistoklis Prodromakis)

A particular challenge is the co-existence of unipolar/bipolar switching and the volatility effects that solid-state resistive memories exhibit. The latest work on how to turn the above mentioned aspects to our benefit in designing cognitive computing based on nanoelectronics will be discussed.

Cognitive Computing with Emerging Nanodevices: Material point of view (by Doo Seok Jeong)

Nanoionics-based electronic transport behavior and related electronic devices, anion- and cation-migration-induced resistive memories, popularly referred to as resistive random access memories will be discussed. In particular, nanoionics thin film materials for artificial neurons and chemical synapses, and implementation of neural functionalities such as spike firing, short- and long-term memories at the single nerve cell level by means of point-defect-migration-dynamics in nanoionic systems will be explained.

Brain-Inspired Computing with Synaptic Devices (by Shimeng Yu)

An overview of the design considerations of synaptic device, benchmark the synaptic device candidates such as PCM, RRAM, CBRAM will be discussed. The target applications such as visual/auditory data processing using devices that consume sub-pJ energy per spike will also be demonstrated.

Introducing resistive memory devices to the SKIM architecture and potentials for harnessing device variation for computing (by Jens Burger)

An introduction to the application of emerging nonvolatile memory technologies in computing (by Omid Kavehei)

Recent advancements on ReRAM technology will be highlighted. Resistive memory classification, modeling and applications will be discussed. Variation sources such as temporal -switching cycle to switching cycle- and spatial -device to device- as well as technical information on device characterization, endurance and retention measurements will be discussed. Multi-level cell (multi-bit memory cells), integration and architectural issues of hybrid CMOS-two terminal nanodevices systems will be highlighted. Issues with selection devices and possible solutions will be presented. Experimental results on fixed and plastic weight computation will also be presented.