Transcript Neuromorphic Engineering

Neuromorphic Engineering
Definitions
Carver Mead introduce the term Neuromorphic Engineering to describe:
A new field of engineering whose design principles and architecture
are biologically Inspired.
Neuro “ to do with neurons i.e. neurally inspired”
Morphic “ structure or form”
Emulates the functional structure of neurobiological systems.
Principles of neuromorphic technology
Build machines that have similar perception capabilities as human perception
Adaptable and self organising
Robust to changing environments
Realisation of future “THINKING” machines
(intelligent and interactive systems)
Why Neuromorphic Engineering?
Interest in exploring
neuroscience
Interest in building
neurally inspired systems
Key Advantages
 emulating biological systems in real time
 attempt to replicate the computational power of brain effectively
 What if our primitive gates were a neuron computation? a synapse
computation? a piece of dendritic cable?
 Efficient implementations compute in their memory elements more efficient
than directly reading all the coefficients
 Precise systems out of imprecise parts
What does neuromorphic Engineering
involve?
Analyze
neurophysiological
functions in order to
reproduce neuronal
structures and
architectures
Build neuronal
networks and rely on
modelers for
explanation and
modeling of natural
phenomena
Brain vs. Computer
Biological brains and digital computers are both complex information
processing systems. But here the similarities end
Energy Efficiency
Progress of electronic information processing over past 60 years:
 dramatic improvements:
 from 5 Joules / instruction (vacuum tube computer, 1940s)
 to 0.0000000001 Joules / instruction (ARM968)
 50,000,000,000 times better
 Raw performance increase about 1 million
Energy efficiency
 Chip: 10-11 J/operation
 Computer system level: 10-9 J/operation
 Brain: 10-15J/operation
 Brain is 1 million times more energy efficient!!!
Computing Power: Human Brain vs.
Computer




Massive parallelism (1011 neurons)
Massive connectivity (1015 synapses)
Low-speed components (~1 – 100 Hz)
>1016 complex operations / second
(10 Petaflops!!!)
 10-15 watts!!!
 1.5 kg
Applications of
Neuromorphic Systems
COMPUTER
SCIENCE
ELECTRICAL
ENGINEERING
NEUROMORPHIC
ENGINEERING
NEUROSCIENCE
- Sensory systems
- Biorobots
- Neuron modelling
- Unsupervised learning
- Pattern understanding
Neuromorphic systems
Silicon Retina
Learning and adaptation
silicon systems
Koala-obstacle/tracking robot
Silicon Cochlea
Electronic Nose
•“Sniff out” odors
•Chemical sensors
•Drug traffic control
•Bio terror detection
Audio Systems
•Audio front ends
•Signal processing systems
•Hearing aids
•Cochlear implants
Spiking camera
Sensorimotor
Systems
•Intelligent robotics
•Intelligent controls
•Locomotive
systems
Biology-Inspired “Neuromorphic” Vision
Biological Paradigm
 Very successful branch of neuromorphic engineering:
sensory transduction  vision
 Neuromorphic vision sensors sense and process visual
information in a pixel-level manner
Functional Model
Electrical Model
VLSI Design
Neuromorphic Vision
Sensor
Neural Disorder Control Parkinson’s
disease Seizure prediction and control
Research works: Blue brain
IBM developing the “Blue brain”
IBM, in partnership with scientists at Switzerland’s Ecole
Polytechnique Federale De Lausanne’s(EPFL) Brain and Mind
Institute will begin simulating the brain’s biological systems.
Research works: SpiNNaker
SpiNNaker is a novel massively-parallel computer architecture,
inspired by the fundamental structure and function of the human
brain, which itself is composed of billions of simple computing
elements, communicating using unreliable spikes.
Research works: Brain Scale project
BrainScale (Brain-inspired multiscale computation in neuromorphic
hybrid systems) was an EU FET-Proactive FP7 funded research
project. The project started on 1 January 2011 and ended on 31 March
2015. It was a collaboration of 19 research groups from 10 European
countries. The hardware development on the neuromorphic computing
systems is continued in the Human Brain Project (HBP) in the
Neuromorphic Computing Platform.
Research works: Neurogrid
Neurogrid is a multi-chip system developed by Kwabena Boahen and
his group at Stanford University. Objective is to emulate neurons
 Composed of a 4x4 array of Neurocores
 Each Neurocore contains a 256x256 array of neuron circuits with
up to 6,000 synapse connections
Research works: FACETS Project
 Fast Analog Computing with
Emergent Transient States
(FACETS)
 A project designed by an
international collective of
scientists and engineers
funded by the European
Union
 Recently developed a chip
containing 200,000 neuron
circuits connected by 50
million synapses.
Research works: MIT silicon synapse
Researchers at the Massachusetts Institute of Technology have
made a huge leap forward with a new chip that mimics the way
the neurons of the brain interact with one another.
Neuromorphic Chips
Future : Silicon Cognition
Future of neuromorphic systems
Implantable medical electronics
Increased human computer interaction
Intelligent transportantion systems
Learning, pattern recognition
Robot control(self motion estimation)
Learning higher order perceptual computation
Brain Machine Interface (BMI)
 A brain-machine interface is a direct communication
pathway between a human or animal brain ( or brain
cell culture) and an external device.
 Sometimes called a direct neural interface or a brain
computer interface (BCI).
Motivation for BMI Research
 In USA, more than
200,000 patients live with
the motor sequelae
(consequences) of serious
injury. There are two ways
to help them restore some
motor function:
 Repair the damaged nerve
axons.
 Build neuroprosthetic
device.
Using BCI
Typing words by mind
Help impaired hands to
grasp by mind
Play videogames by mind
Structure of the bidirectional BMI
‫‪Online Calibration Process‬‬
‫‪ . 1‬تولید دیتای کالیبراسیون توسط برنامه ‪M1-S1‬‬
‫‪. 2‬دریافت دیتای تولید شده و چیدن آن در ماتریس داده ها‬
‫‪ . 3‬قرار دادن دیتای تولید شده در برنامه کالیبراسیون‬
‫‪. 4‬تولید ماتریس )‪ D(distance‬و ‪ Force‬و مشخص شدن ناحیه بندی دیتاها برای داده های کالیبراسیون جهت‬
‫استفاده در متن برنامه اصلی‬
‫‪Offline Test Process‬‬
‫‪ .1‬دریافت داده از برنامه‬
‫‪M1-S1‬‬
‫‪ .2‬محاسبه ماتریس ‪ D‬برای این‬
‫داده با استفاده از داده های‬
‫کالیبراسیون‬
‫‪ .7‬اعمال این دیتای انکد شده‬
‫به برنامه ‪ S1-M1‬جهت‬
‫دریافت داده جدید‬
‫‪ .3‬دیکد کردن این داده با‬
‫استفاده از ماتریس ‪ D‬بدست‬
‫آمده و مشخص نمودن ناحیه‬
‫آن‬
‫‪ .6‬انکد کردن این موقعیت‬
‫جدید جهت اعمال به برنامه‬
‫‪S1-M1‬‬
‫‪ .4‬محاسبه ‪ Force‬با استفده‬
‫از ناحیه دیکد شده با استفاده‬
‫از ماتریس ‪ Force‬تولد شده‬
‫در کالیبراسیون‬
‫‪ .5‬عمال ‪ Force‬تولید شده و‬
‫محاسبه موقعیت جدید جهت‬
‫مشخص نمودن تحریک جدید‬
Simulation Results
Simulation Results
Simulation Results
D=2.31
Simulation Results
Simulation Results
D=20.7976
Simulation Results
D=3.2264
Advantages
 BCIs will help creating a Direct communication pathway
between a human or animal brain and any external devices like
computers.
 BCI has increased the possibility of treatment of disabilities
related to nervous system along with the old technique of
Neuroprosthetics.
 Techniques like EEG, MEG and neurochips have come into
discussions since the BCI application have started developing.
 This has provided a new work area for scientists and researchers
around the world.
Disadvantages
 In case of Invasive BCI there is a risk of formation of scar
tissue.
 There is a need of extensive training before user can use
techniques like EEG
 BCI techniques still require much enhancement before they can
be used by users as they are slow.
 Ethical implications of BCI will arise in future
 BCI techniques are costly. It requires a lot of money to set up
the BCI environment.
Our Publications




Ranjbar, M., & Amiri, M. (2015). An analog astrocyte–neuron
interaction circuit for neuromorphic applications. Journal of
Computational Electronics, 14(3), 694-706.
Ranjbar, M., & Amiri, M. (2015). Analog implementation of
neuron–astrocyte interaction in tripartite synapse. Journal of
Computational Electronics, 1-13.
Piri, M., Amiri, M., & Amiri, M. (2015). A bio-inspired stimulator
to desynchronize epileptic cortical population models: A digital
implementation framework. Neural Networks, 67, 74-83.
Nazari, S., Faez, K., Karami, E., & Amiri, M. (2014). A digital
neurmorphic circuit for a simplified model of astrocyte dynamics.
Neuroscience letters, 582, 21-26.





Nazari, S., Faez, K., Amiri, M., & Karami, E. (2015). A novel
digital implementation of neuron–astrocyte interactions. Journal of
Computational Electronics, 14(1), 227-239.
Nazari, S., Faez, K., Amiri, M., & Karami, E. (2015). A digital
implementation of neuron–astrocyte interaction for neuromorphic
applications. Neural Networks, 66, 79-90.
Nazari, S., Amiri, M., Faez, K., & Amiri, M. (2015). Multiplierless digital implementation of neuron–astrocyte signalling on
FPGA. Neurocomputing.
Nazari, S., Faez, K., & Amiri, M. A multiplier-less digital design of
a bio-inspired stimulator to suppress synchronized regime in a
large-scale, sparsely connected neural network. Neural Computing
and Applications, 1-16.
Ranjbar, M., & Amiri, M. On the role of astrocyte analog circuit in
neural frequency adaptation. Neural Computing and Applications,
1-13.
THANK YOU!