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

Chip data illustrating contrast enhancement. (A) An odor with a constant concentration was presented at the 200-ms time point, and persisted till the 500-ms time point (solid black line). As an illustration, the spiking activity (vertical bars) of seven mitral cells in response to the odor is shown. The mitral cells receiving input from OSNs that most strongly respond to the odor exhibit increased activity whereas those receiving input from weakly responsive OSNs are inhibited below baseline, such that they are less active even than completely insensitive mitral cells (see Figure 9). The activity patterns of the mitral cells shown are presented in a one-dimensional, two-tailed chemotopic arrangement for illustrative purposes; see Cleland and Linster (2012) for a high-dimensional illustration. The mitral cell depicted in the fourth row received input from the OSNs that responded most strongly, whereas the mitral cells depicted in the rows above and below it received progressively weaker OSN activity. The result is a sharply tuned population tuning curve (in blue) with a distinct inhibitory surround (the spike count is for the duration of the stimulus). (B) The representation of two different odors (two peaks) by a population of mitral cells (distributed along the abscissa in a one-dimensional chemotopic ordering for illustrative purposes). Non-topographical feed-forward (PGo) inhibition generates an inhibitory surround and reduces the overlap of the two combinatorial odor representations (solid line) compared to the overlap that would occur in the absence of PGo inhibition (dashed line).
© Copyright Policy - open-access
Related In: Results  -  Collection

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

Figure 8: Chip data illustrating contrast enhancement. (A) An odor with a constant concentration was presented at the 200-ms time point, and persisted till the 500-ms time point (solid black line). As an illustration, the spiking activity (vertical bars) of seven mitral cells in response to the odor is shown. The mitral cells receiving input from OSNs that most strongly respond to the odor exhibit increased activity whereas those receiving input from weakly responsive OSNs are inhibited below baseline, such that they are less active even than completely insensitive mitral cells (see Figure 9). The activity patterns of the mitral cells shown are presented in a one-dimensional, two-tailed chemotopic arrangement for illustrative purposes; see Cleland and Linster (2012) for a high-dimensional illustration. The mitral cell depicted in the fourth row received input from the OSNs that responded most strongly, whereas the mitral cells depicted in the rows above and below it received progressively weaker OSN activity. The result is a sharply tuned population tuning curve (in blue) with a distinct inhibitory surround (the spike count is for the duration of the stimulus). (B) The representation of two different odors (two peaks) by a population of mitral cells (distributed along the abscissa in a one-dimensional chemotopic ordering for illustrative purposes). Non-topographical feed-forward (PGo) inhibition generates an inhibitory surround and reduces the overlap of the two combinatorial odor representations (solid line) compared to the overlap that would occur in the absence of PGo inhibition (dashed line).

Mentions: The connections between PGo and mitral cells (see Figure 1) sharpen the secondary representations of odors as described in Section “Computations of the Olfactory Bulb Glomerular Layer.” Mitral cell spike patterns measured from the chip confirm that feed-forward inhibition by PGo cells indeed can yield output that is more precisely tuned to specific odorants than are the broad primary representations mediated by OSNs (sensors). This chemotopic contrast enhancement is illustrated in Figure 8A. We represented the presence of an odor by externally driving OSN inputs such that a few had high spike rates (depicting strongly responsive chemosensors), some had moderate spike rates (depicting moderately/weakly responsive chemosensors), and some did not increase their spiking over the background rate (depicting insensitive chemosensors). As shown in the figure, those mitral cells corresponding to the most strongly activated OSNs increased their spiking activity, while those corresponding to the moderately/weakly activated OSNs shut down, resulting in a sharp population tuning curve. As illustrated in Figure 8B, this reduction in the overlap of the combinatorial representations of similar analytes creates a sharply decorrelated set of odor representations. The basic decorrelation function is illustrated in Figure 9.


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)

Chip data illustrating contrast enhancement. (A) An odor with a constant concentration was presented at the 200-ms time point, and persisted till the 500-ms time point (solid black line). As an illustration, the spiking activity (vertical bars) of seven mitral cells in response to the odor is shown. The mitral cells receiving input from OSNs that most strongly respond to the odor exhibit increased activity whereas those receiving input from weakly responsive OSNs are inhibited below baseline, such that they are less active even than completely insensitive mitral cells (see Figure 9). The activity patterns of the mitral cells shown are presented in a one-dimensional, two-tailed chemotopic arrangement for illustrative purposes; see Cleland and Linster (2012) for a high-dimensional illustration. The mitral cell depicted in the fourth row received input from the OSNs that responded most strongly, whereas the mitral cells depicted in the rows above and below it received progressively weaker OSN activity. The result is a sharply tuned population tuning curve (in blue) with a distinct inhibitory surround (the spike count is for the duration of the stimulus). (B) The representation of two different odors (two peaks) by a population of mitral cells (distributed along the abscissa in a one-dimensional chemotopic ordering for illustrative purposes). Non-topographical feed-forward (PGo) inhibition generates an inhibitory surround and reduces the overlap of the two combinatorial odor representations (solid line) compared to the overlap that would occur in the absence of PGo inhibition (dashed line).
© Copyright Policy - open-access
Related In: Results  -  Collection

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

Figure 8: Chip data illustrating contrast enhancement. (A) An odor with a constant concentration was presented at the 200-ms time point, and persisted till the 500-ms time point (solid black line). As an illustration, the spiking activity (vertical bars) of seven mitral cells in response to the odor is shown. The mitral cells receiving input from OSNs that most strongly respond to the odor exhibit increased activity whereas those receiving input from weakly responsive OSNs are inhibited below baseline, such that they are less active even than completely insensitive mitral cells (see Figure 9). The activity patterns of the mitral cells shown are presented in a one-dimensional, two-tailed chemotopic arrangement for illustrative purposes; see Cleland and Linster (2012) for a high-dimensional illustration. The mitral cell depicted in the fourth row received input from the OSNs that responded most strongly, whereas the mitral cells depicted in the rows above and below it received progressively weaker OSN activity. The result is a sharply tuned population tuning curve (in blue) with a distinct inhibitory surround (the spike count is for the duration of the stimulus). (B) The representation of two different odors (two peaks) by a population of mitral cells (distributed along the abscissa in a one-dimensional chemotopic ordering for illustrative purposes). Non-topographical feed-forward (PGo) inhibition generates an inhibitory surround and reduces the overlap of the two combinatorial odor representations (solid line) compared to the overlap that would occur in the absence of PGo inhibition (dashed line).
Mentions: The connections between PGo and mitral cells (see Figure 1) sharpen the secondary representations of odors as described in Section “Computations of the Olfactory Bulb Glomerular Layer.” Mitral cell spike patterns measured from the chip confirm that feed-forward inhibition by PGo cells indeed can yield output that is more precisely tuned to specific odorants than are the broad primary representations mediated by OSNs (sensors). This chemotopic contrast enhancement is illustrated in Figure 8A. We represented the presence of an odor by externally driving OSN inputs such that a few had high spike rates (depicting strongly responsive chemosensors), some had moderate spike rates (depicting moderately/weakly responsive chemosensors), and some did not increase their spiking over the background rate (depicting insensitive chemosensors). As shown in the figure, those mitral cells corresponding to the most strongly activated OSNs increased their spiking activity, while those corresponding to the moderately/weakly activated OSNs shut down, resulting in a sharp population tuning curve. As illustrated in Figure 8B, this reduction in the overlap of the combinatorial representations of similar analytes creates a sharply decorrelated set of odor representations. The basic decorrelation function is illustrated in Figure 9.

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