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

(A) With the ET/sSA network off, mitral cell population activity increases monotonically in proportion to input intensity (odor concentration, [C]). The spikes of those mitral cells that are most sharply tuned to the presented odor are depicted in red and indicated by red arrows; additional, weakly activated mitral cells also are recruited in to the active ensemble as concentration increases. With the ET/sSA network on, aggregate mitral cell population activity remains largely independent of concentration. Odor identity is represented by the profile of the most sharply tuned mitral cells, and their relative activities remain essentially stable after normalization. (B) Quantification of the normalizing effect of the ET/sSA network. Spiking activity of mitral cells that are moderately or poorly tuned to a given stimulus (i.e., the count of all black marks in the raster of (A), excluding the spikes from sharply tuned neurons as denoted in red) was plotted against odor concentration. The inhibitory effects of the ET/sSA network limit or prevent firing in more weakly tuned mitral cells, such as those that are recruited only at higher odor concentrations. At high concentration the normalizing effects reduce the spike count by as much as 80%. (C) Effects of ET/sSA network normalization on contrast enhancement. Contrast was calculated as the total number of spikes in the most strongly tuned neurons divided by the total number of spikes in the population. With the ET/sSA network disconnected (top), the recruitment of weakly tuned mitral cells by elevated odor concentrations prevents improvements in contrast across mitral cell population activity. With the ET/sSA network active (bottom), increases in concentration result in net inhibition of marginally activated neurons, increasing the contrast between them and the most strongly activated neurons.
© Copyright Policy - open-access
Related In: Results  -  Collection

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

Figure 10: (A) With the ET/sSA network off, mitral cell population activity increases monotonically in proportion to input intensity (odor concentration, [C]). The spikes of those mitral cells that are most sharply tuned to the presented odor are depicted in red and indicated by red arrows; additional, weakly activated mitral cells also are recruited in to the active ensemble as concentration increases. With the ET/sSA network on, aggregate mitral cell population activity remains largely independent of concentration. Odor identity is represented by the profile of the most sharply tuned mitral cells, and their relative activities remain essentially stable after normalization. (B) Quantification of the normalizing effect of the ET/sSA network. Spiking activity of mitral cells that are moderately or poorly tuned to a given stimulus (i.e., the count of all black marks in the raster of (A), excluding the spikes from sharply tuned neurons as denoted in red) was plotted against odor concentration. The inhibitory effects of the ET/sSA network limit or prevent firing in more weakly tuned mitral cells, such as those that are recruited only at higher odor concentrations. At high concentration the normalizing effects reduce the spike count by as much as 80%. (C) Effects of ET/sSA network normalization on contrast enhancement. Contrast was calculated as the total number of spikes in the most strongly tuned neurons divided by the total number of spikes in the population. With the ET/sSA network disconnected (top), the recruitment of weakly tuned mitral cells by elevated odor concentrations prevents improvements in contrast across mitral cell population activity. With the ET/sSA network active (bottom), increases in concentration result in net inhibition of marginally activated neurons, increasing the contrast between them and the most strongly activated neurons.

Mentions: Figure 10 illustrates the normalizing effects of the ET/sSA cell network on the activity of mitral cells in the chip. The uniform level of activation integrated by this network excites PGe and ET cells in all glomeruli and thereby uniformly inhibits all mitral cells in proportion to the global average level of OSN activation. This process generates a relational representation among mitral cells in which the relative (rather than the absolute) pattern of activity best represents the identity of an odor and also helps to preserve recognizable odor-specific activity patterns across concentrations (concentration invariance). This normalization also is necessary for NTCE to function across different concentrations. Optimal concentration invariance still will be limited by sensor non-linearities and ultimately may be improved by an adaptive algorithm as discussed in Cleland et al. (2012).


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)

(A) With the ET/sSA network off, mitral cell population activity increases monotonically in proportion to input intensity (odor concentration, [C]). The spikes of those mitral cells that are most sharply tuned to the presented odor are depicted in red and indicated by red arrows; additional, weakly activated mitral cells also are recruited in to the active ensemble as concentration increases. With the ET/sSA network on, aggregate mitral cell population activity remains largely independent of concentration. Odor identity is represented by the profile of the most sharply tuned mitral cells, and their relative activities remain essentially stable after normalization. (B) Quantification of the normalizing effect of the ET/sSA network. Spiking activity of mitral cells that are moderately or poorly tuned to a given stimulus (i.e., the count of all black marks in the raster of (A), excluding the spikes from sharply tuned neurons as denoted in red) was plotted against odor concentration. The inhibitory effects of the ET/sSA network limit or prevent firing in more weakly tuned mitral cells, such as those that are recruited only at higher odor concentrations. At high concentration the normalizing effects reduce the spike count by as much as 80%. (C) Effects of ET/sSA network normalization on contrast enhancement. Contrast was calculated as the total number of spikes in the most strongly tuned neurons divided by the total number of spikes in the population. With the ET/sSA network disconnected (top), the recruitment of weakly tuned mitral cells by elevated odor concentrations prevents improvements in contrast across mitral cell population activity. With the ET/sSA network active (bottom), increases in concentration result in net inhibition of marginally activated neurons, increasing the contrast between them and the most strongly activated neurons.
© Copyright Policy - open-access
Related In: Results  -  Collection

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

Figure 10: (A) With the ET/sSA network off, mitral cell population activity increases monotonically in proportion to input intensity (odor concentration, [C]). The spikes of those mitral cells that are most sharply tuned to the presented odor are depicted in red and indicated by red arrows; additional, weakly activated mitral cells also are recruited in to the active ensemble as concentration increases. With the ET/sSA network on, aggregate mitral cell population activity remains largely independent of concentration. Odor identity is represented by the profile of the most sharply tuned mitral cells, and their relative activities remain essentially stable after normalization. (B) Quantification of the normalizing effect of the ET/sSA network. Spiking activity of mitral cells that are moderately or poorly tuned to a given stimulus (i.e., the count of all black marks in the raster of (A), excluding the spikes from sharply tuned neurons as denoted in red) was plotted against odor concentration. The inhibitory effects of the ET/sSA network limit or prevent firing in more weakly tuned mitral cells, such as those that are recruited only at higher odor concentrations. At high concentration the normalizing effects reduce the spike count by as much as 80%. (C) Effects of ET/sSA network normalization on contrast enhancement. Contrast was calculated as the total number of spikes in the most strongly tuned neurons divided by the total number of spikes in the population. With the ET/sSA network disconnected (top), the recruitment of weakly tuned mitral cells by elevated odor concentrations prevents improvements in contrast across mitral cell population activity. With the ET/sSA network active (bottom), increases in concentration result in net inhibition of marginally activated neurons, increasing the contrast between them and the most strongly activated neurons.
Mentions: Figure 10 illustrates the normalizing effects of the ET/sSA cell network on the activity of mitral cells in the chip. The uniform level of activation integrated by this network excites PGe and ET cells in all glomeruli and thereby uniformly inhibits all mitral cells in proportion to the global average level of OSN activation. This process generates a relational representation among mitral cells in which the relative (rather than the absolute) pattern of activity best represents the identity of an odor and also helps to preserve recognizable odor-specific activity patterns across concentrations (concentration invariance). This normalization also is necessary for NTCE to function across different concentrations. Optimal concentration invariance still will be limited by sensor non-linearities and ultimately may be improved by an adaptive algorithm as discussed in Cleland et al. (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