Presentation on theme: "Neuromorphic Engineering"— Presentation transcript:
1 Neuromorphic Engineering University of OxfordDepartment of Engineering ScienceNatasha Chia
2 Outline Definition of neuromorphic systems Principles of neuromorphic technologyTypical applications of neuromorphic technologyThe basic computational elementNeural codes and information representationFuture of Neuromorphic systems
3 Emulates the functional structure of neurobiological systems. DefinitionsCarver Mead introducec the term Neuromorphic Engineering to describeA new field of engineering whose design principles and architecture are biologicallyInspired.Neuro “ to do with neurons i.e. neurally inspired”Morphic “ structure or form”Emulates the functional structure of neurobiological systems.Neuromorphic systems are systems implemented on silicon,whose architecture and design are based on neurobiology
4 Principles of neuromorphic technology Build machines that have similar perception capabilities as human perceptionAdaptable and self organisingRobust to changing environmentsRealisation of future “THINKING” machines(intelligent and interactive systems)These understanding are creating new opportunities for intelligent systems designProvides promising methods to tackle a wide variety of design related problems
7 Neuromorphic systems Silicon Retina Learning and adaptation silicon systemsKoala-obstacle/tracking robotSilicon CochleaExisting app are modest and the challenge of designing fully autonomous systems at the levels achieved by biological systemslies ahead. These devices are mostly analog and used as front end sensorsTouchPad
8 Example of Modelling approaches Functional approach adopted(The goal is to construct a biologically constrained framework for speech recognition at the network level)so functional components are used in the models not actual biophysical components.Build a hierarchical speech system based on a model of the mammalian auditory systemThe undertaking is to construct a biologically constrained framework for speech recognition at the network level.Parallel distributed analog correlation based processing is the basis of VLSI systems that emulate the function of neural information processing in biological systems
9 The Basic computational element : The Neuron A typical neuron accepts input from its large dendritic tree and combines the information to produce a single output at its axon.
11 History of neuron modelling 1943 McCulloch & Pits1963 Hodgkin & Huxley modelled axon of giant squid.1970 Kiang & Gerstein numerical analysis of interactions in nerve cells1983 Cohan & Mpitson Deterministic chaos used to describe behaviour of single neurons1985 Meyer et al Discovery of electrical interaction between neurons1992 Bower Quantitative evaluation of functional data1993 Knopf and Gupta Fundamental neural processing element1995 Bressler Parallel processing of information
12 Neural code and information representation How does spikes represent sensory information?Models of neuron spiking mechanismStimulus Response featuresGroup behaviour
13 Stimulus Response features Average firing ratePosition of eachneuronal dischargeInstantaneous firingprobability.(A ganglion’s firing rate depends on stimulation)Spike trains can only be encoded under two ideal conditions1.Spike times must be precise2.Distance between two events must be more than the spike time-jitter.
14 Spiking Neuron models Integrate and fire Stimulus Response Model (SRM) Neurons spike regularly in responseto an external current. Rate ofspiking increases with the magnitudeof the stimulus currentStimulus Response Model (SRM)The stimulus is generalized into certain patterns in order to quantify the stimulus response
15 Neural codingInformation is conveyed to the brain in parallel by spike trains of the nerve fibers.
16 Group behaviour of spikes One of the challenges has been to understand the relative contribution of various groups of spikesOscillatory coupling effects and amplitude dynamics in two or more populations of neurons will be important topics for future researchSome other examples: STDP(Spike Time-dependent plasticity)Group neurons with correlated inputsBrief talk on why the gp behaviour of spikes is not well understood
17 Future of neuromorphic systems Implantable medical electronicsIncreased human computer interactionIntelligent transportantion systemsLearning, pattern recognitionRobot control(self motion estimation)Learning higher order perceptual computationNeu en is still an evolving field and has not matured into an established discipline.Human computer interaction is shifted away from the keyboard monitor to other natural means of interaction s a speech.Neural based computation(learning,recognition and attention