Limits...
Implementation of olfactory bulb glomerular-layer computations in a digital neurosynaptic core.

Imam N, Cleland TA, Manohar R, Merolla PA, Arthur JV, Akopyan F, Modha DS - Front Neurosci (2012)

Bottom Line: Our system is based on a digital neuromorphic chip consisting of 256 leaky-integrate-and-fire neurons, 1024 × 256 crossbar synapses, and address-event representation communication circuits.The neural circuits configured in the chip reflect established connections among mitral cells, periglomerular cells, external tufted cells, and superficial short-axon cells within the olfactory bulb, and accept input from convergent sets of sensors configured as olfactory sensory neurons.Our circuits, consuming only 45 pJ of active power per spike with a power supply of 0.85 V, can be used as the first stage of processing in low-power artificial chemical sensing devices inspired by natural olfactory systems.

View Article: PubMed Central - PubMed

Affiliation: Computer Systems Lab, Department of Electrical and Computer Engineering, Cornell University Ithaca, NY, USA.

ABSTRACT
We present a biomimetic system that captures essential functional properties of the glomerular layer of the mammalian olfactory bulb, specifically including its capacity to decorrelate similar odor representations without foreknowledge of the statistical distributions of analyte features. Our system is based on a digital neuromorphic chip consisting of 256 leaky-integrate-and-fire neurons, 1024 × 256 crossbar synapses, and address-event representation communication circuits. The neural circuits configured in the chip reflect established connections among mitral cells, periglomerular cells, external tufted cells, and superficial short-axon cells within the olfactory bulb, and accept input from convergent sets of sensors configured as olfactory sensory neurons. This configuration generates functional transformations comparable to those observed in the glomerular layer of the mammalian olfactory bulb. Our circuits, consuming only 45 pJ of active power per spike with a power supply of 0.85 V, can be used as the first stage of processing in low-power artificial chemical sensing devices inspired by natural olfactory systems.

No MeSH data available.


Related in: MedlinePlus

Mitral cell population activity in the chip in response to the presentation of four different odors. This activity constitutes the final output of the system in its present configuration. Spiking patterns from 24 of the 48 mitral cells on the chip are depicted. Odorant concentration [C] was modulated as depicted at the bottom of the figure. Compared to the primary representation of the chemical environment by the broadly tuned, concentration-sensitive population of simulated sensors driving the chip, this secondary representation is more sharply tuned, normalized, and exhibits an increased signal-to-noise ratio. For final odor identification and discrimination, this output activity can be sent to a standard pattern recognition engine, or to biomimetic circuits representing higher processing stages in olfactory pathways.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC3368244&req=5

Figure 12: Mitral cell population activity in the chip in response to the presentation of four different odors. This activity constitutes the final output of the system in its present configuration. Spiking patterns from 24 of the 48 mitral cells on the chip are depicted. Odorant concentration [C] was modulated as depicted at the bottom of the figure. Compared to the primary representation of the chemical environment by the broadly tuned, concentration-sensitive population of simulated sensors driving the chip, this secondary representation is more sharply tuned, normalized, and exhibits an increased signal-to-noise ratio. For final odor identification and discrimination, this output activity can be sent to a standard pattern recognition engine, or to biomimetic circuits representing higher processing stages in olfactory pathways.

Mentions: As a result of the transformations in the chip, the secondary representation generated across mitral cell spike trains exhibits an improved signal-to-noise ratio, and is sharply tuned and normalized with respect to the global activity. Output representations from the chip (illustrated in Figure 12) can be further processed, analyzed, and/or classified by standard pattern recognition engines running in parallel with our system, or by another neuromorphic circuit (e.g., based on a new configuration of the present system) that replicates some functions of higher processing centers in biological olfactory pathways, such as the perceptual learning and higher-order feedback mechanisms exhibited by the in vivo olfactory system. Importantly, such neuromorphic implementations of subsequent layers of olfactory processing (e.g., the external plexiform layer of the olfactory bulb) are likely to require adaptive plasticity in synaptic weights and/or neuronal properties. The experience-dependent transformation of odor representation is likely to be a powerful technique for selectively grouping regions of olfactory variance into “odors” in accordance with the similarities of their implications. Such dynamic remapping of the connections within the chip (Sun Seo et al., 2011) can be achieved by enabling the neurons to change the state of the crossbar during run-time based on plasticity rules such as spike timing-dependent plasticity (Dan and Poo, 2006; Linster and Cleland, 2010).


Implementation of olfactory bulb glomerular-layer computations in a digital neurosynaptic core.

Imam N, Cleland TA, Manohar R, Merolla PA, Arthur JV, Akopyan F, Modha DS - Front Neurosci (2012)

Mitral cell population activity in the chip in response to the presentation of four different odors. This activity constitutes the final output of the system in its present configuration. Spiking patterns from 24 of the 48 mitral cells on the chip are depicted. Odorant concentration [C] was modulated as depicted at the bottom of the figure. Compared to the primary representation of the chemical environment by the broadly tuned, concentration-sensitive population of simulated sensors driving the chip, this secondary representation is more sharply tuned, normalized, and exhibits an increased signal-to-noise ratio. For final odor identification and discrimination, this output activity can be sent to a standard pattern recognition engine, or to biomimetic circuits representing higher processing stages in olfactory pathways.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC3368244&req=5

Figure 12: Mitral cell population activity in the chip in response to the presentation of four different odors. This activity constitutes the final output of the system in its present configuration. Spiking patterns from 24 of the 48 mitral cells on the chip are depicted. Odorant concentration [C] was modulated as depicted at the bottom of the figure. Compared to the primary representation of the chemical environment by the broadly tuned, concentration-sensitive population of simulated sensors driving the chip, this secondary representation is more sharply tuned, normalized, and exhibits an increased signal-to-noise ratio. For final odor identification and discrimination, this output activity can be sent to a standard pattern recognition engine, or to biomimetic circuits representing higher processing stages in olfactory pathways.
Mentions: As a result of the transformations in the chip, the secondary representation generated across mitral cell spike trains exhibits an improved signal-to-noise ratio, and is sharply tuned and normalized with respect to the global activity. Output representations from the chip (illustrated in Figure 12) can be further processed, analyzed, and/or classified by standard pattern recognition engines running in parallel with our system, or by another neuromorphic circuit (e.g., based on a new configuration of the present system) that replicates some functions of higher processing centers in biological olfactory pathways, such as the perceptual learning and higher-order feedback mechanisms exhibited by the in vivo olfactory system. Importantly, such neuromorphic implementations of subsequent layers of olfactory processing (e.g., the external plexiform layer of the olfactory bulb) are likely to require adaptive plasticity in synaptic weights and/or neuronal properties. The experience-dependent transformation of odor representation is likely to be a powerful technique for selectively grouping regions of olfactory variance into “odors” in accordance with the similarities of their implications. Such dynamic remapping of the connections within the chip (Sun Seo et al., 2011) can be achieved by enabling the neurons to change the state of the crossbar during run-time based on plasticity rules such as spike timing-dependent plasticity (Dan and Poo, 2006; Linster and Cleland, 2010).

Bottom Line: Our system is based on a digital neuromorphic chip consisting of 256 leaky-integrate-and-fire neurons, 1024 × 256 crossbar synapses, and address-event representation communication circuits.The neural circuits configured in the chip reflect established connections among mitral cells, periglomerular cells, external tufted cells, and superficial short-axon cells within the olfactory bulb, and accept input from convergent sets of sensors configured as olfactory sensory neurons.Our circuits, consuming only 45 pJ of active power per spike with a power supply of 0.85 V, can be used as the first stage of processing in low-power artificial chemical sensing devices inspired by natural olfactory systems.

View Article: PubMed Central - PubMed

Affiliation: Computer Systems Lab, Department of Electrical and Computer Engineering, Cornell University Ithaca, NY, USA.

ABSTRACT
We present a biomimetic system that captures essential functional properties of the glomerular layer of the mammalian olfactory bulb, specifically including its capacity to decorrelate similar odor representations without foreknowledge of the statistical distributions of analyte features. Our system is based on a digital neuromorphic chip consisting of 256 leaky-integrate-and-fire neurons, 1024 × 256 crossbar synapses, and address-event representation communication circuits. The neural circuits configured in the chip reflect established connections among mitral cells, periglomerular cells, external tufted cells, and superficial short-axon cells within the olfactory bulb, and accept input from convergent sets of sensors configured as olfactory sensory neurons. This configuration generates functional transformations comparable to those observed in the glomerular layer of the mammalian olfactory bulb. Our circuits, consuming only 45 pJ of active power per spike with a power supply of 0.85 V, can be used as the first stage of processing in low-power artificial chemical sensing devices inspired by natural olfactory systems.

No MeSH data available.


Related in: MedlinePlus