Limits...
Nonlinear circuits for naturalistic visual motion estimation.

Fitzgerald JE, Clark DA - Elife (2015)

Bottom Line: Furthermore, a diversity of inputs to motion detecting neurons can provide access to more complex higher-order correlations.Collectively, these results illustrate how non-canonical computations improve motion estimation with naturalistic inputs.This argues that the complexity of the fly's motion computations, implemented in its elaborate circuits, represents a valuable feature of its visual motion estimator.

View Article: PubMed Central - PubMed

Affiliation: Center for Brain Science, Harvard University, Cambridge, United States.

ABSTRACT
Many animals use visual signals to estimate motion. Canonical models suppose that animals estimate motion by cross-correlating pairs of spatiotemporally separated visual signals, but recent experiments indicate that humans and flies perceive motion from higher-order correlations that signify motion in natural environments. Here we show how biologically plausible processing motifs in neural circuits could be tuned to extract this information. We emphasize how known aspects of Drosophila's visual circuitry could embody this tuning and predict fly behavior. We find that segregating motion signals into ON/OFF channels can enhance estimation accuracy by accounting for natural light/dark asymmetries. Furthermore, a diversity of inputs to motion detecting neurons can provide access to more complex higher-order correlations. Collectively, these results illustrate how non-canonical computations improve motion estimation with naturalistic inputs. This argues that the complexity of the fly's motion computations, implemented in its elaborate circuits, represents a valuable feature of its visual motion estimator.

Show MeSH
The non-multiplicative nonlinearity model can be tuned to account for the positive-negative parity asymmetry in the psychophysical data.In this numerical experiment, we tuned the model nonlinearity to optimize a fit to the psychophysical data. (A) In this case, the tuned model could reproduce the 2-point glider data well. (B) This tuned model could also reproduce the differential amplitudes of the positive and negative parity responses. Thus, the non-multiplicative nonlinearity model repairs an architectural defect of the weighted 4-quadrant model (Figure 3—figure supplement 2).DOI:http://dx.doi.org/10.7554/eLife.09123.011
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4663970&req=5

fig4s1: The non-multiplicative nonlinearity model can be tuned to account for the positive-negative parity asymmetry in the psychophysical data.In this numerical experiment, we tuned the model nonlinearity to optimize a fit to the psychophysical data. (A) In this case, the tuned model could reproduce the 2-point glider data well. (B) This tuned model could also reproduce the differential amplitudes of the positive and negative parity responses. Thus, the non-multiplicative nonlinearity model repairs an architectural defect of the weighted 4-quadrant model (Figure 3—figure supplement 2).DOI:http://dx.doi.org/10.7554/eLife.09123.011

Mentions: The first of these models recasts the HRC and the weighted 4-quadrant model in a more general architecture. This model is the class of mirror anti-symmetric models that apply a 2-dimensional nonlinearity to the low-pass filtered signal from one point in space and the high-pass filtered signal from a neighboring point in space (Figure 4A). Since the observed glider responses indicate that flies use higher-order correlations of both even and odd order, we model this 2-dimensional nonlinearity as a fourth-order polynomial (‘Materials and methods’). The HRC corresponds to the special case of this nonlinearity that multiplies the two inputs (left, Figure 4B). To emphasize how the model class in Figure 4A generalizes the HRC, we refer to it as the non-multiplicative nonlinearity model. In comparison, the weighted 4-quadrant model corresponds to a different nonlinearity that separately scales a pure multiplication in each quadrant of the Cartesian plane. Compared to the HRC, the optimized forms of both the weighted 4-quadrant model and the non-multiplicative nonlinearity model substantially attenuated positive low-pass filtered signals (middle and right, Figure 4B), though the non-multiplicative nonlinearity shows less attenuation. This model architecture provides enough flexibility to generate the glider response pattern (Figure 4—figure supplement 1).10.7554/eLife.09123.010Figure 4.Several biologically motivated generalizations of the motion estimator further improved estimation performance without sacrificing glider responses.


Nonlinear circuits for naturalistic visual motion estimation.

Fitzgerald JE, Clark DA - Elife (2015)

The non-multiplicative nonlinearity model can be tuned to account for the positive-negative parity asymmetry in the psychophysical data.In this numerical experiment, we tuned the model nonlinearity to optimize a fit to the psychophysical data. (A) In this case, the tuned model could reproduce the 2-point glider data well. (B) This tuned model could also reproduce the differential amplitudes of the positive and negative parity responses. Thus, the non-multiplicative nonlinearity model repairs an architectural defect of the weighted 4-quadrant model (Figure 3—figure supplement 2).DOI:http://dx.doi.org/10.7554/eLife.09123.011
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4663970&req=5

fig4s1: The non-multiplicative nonlinearity model can be tuned to account for the positive-negative parity asymmetry in the psychophysical data.In this numerical experiment, we tuned the model nonlinearity to optimize a fit to the psychophysical data. (A) In this case, the tuned model could reproduce the 2-point glider data well. (B) This tuned model could also reproduce the differential amplitudes of the positive and negative parity responses. Thus, the non-multiplicative nonlinearity model repairs an architectural defect of the weighted 4-quadrant model (Figure 3—figure supplement 2).DOI:http://dx.doi.org/10.7554/eLife.09123.011
Mentions: The first of these models recasts the HRC and the weighted 4-quadrant model in a more general architecture. This model is the class of mirror anti-symmetric models that apply a 2-dimensional nonlinearity to the low-pass filtered signal from one point in space and the high-pass filtered signal from a neighboring point in space (Figure 4A). Since the observed glider responses indicate that flies use higher-order correlations of both even and odd order, we model this 2-dimensional nonlinearity as a fourth-order polynomial (‘Materials and methods’). The HRC corresponds to the special case of this nonlinearity that multiplies the two inputs (left, Figure 4B). To emphasize how the model class in Figure 4A generalizes the HRC, we refer to it as the non-multiplicative nonlinearity model. In comparison, the weighted 4-quadrant model corresponds to a different nonlinearity that separately scales a pure multiplication in each quadrant of the Cartesian plane. Compared to the HRC, the optimized forms of both the weighted 4-quadrant model and the non-multiplicative nonlinearity model substantially attenuated positive low-pass filtered signals (middle and right, Figure 4B), though the non-multiplicative nonlinearity shows less attenuation. This model architecture provides enough flexibility to generate the glider response pattern (Figure 4—figure supplement 1).10.7554/eLife.09123.010Figure 4.Several biologically motivated generalizations of the motion estimator further improved estimation performance without sacrificing glider responses.

Bottom Line: Furthermore, a diversity of inputs to motion detecting neurons can provide access to more complex higher-order correlations.Collectively, these results illustrate how non-canonical computations improve motion estimation with naturalistic inputs.This argues that the complexity of the fly's motion computations, implemented in its elaborate circuits, represents a valuable feature of its visual motion estimator.

View Article: PubMed Central - PubMed

Affiliation: Center for Brain Science, Harvard University, Cambridge, United States.

ABSTRACT
Many animals use visual signals to estimate motion. Canonical models suppose that animals estimate motion by cross-correlating pairs of spatiotemporally separated visual signals, but recent experiments indicate that humans and flies perceive motion from higher-order correlations that signify motion in natural environments. Here we show how biologically plausible processing motifs in neural circuits could be tuned to extract this information. We emphasize how known aspects of Drosophila's visual circuitry could embody this tuning and predict fly behavior. We find that segregating motion signals into ON/OFF channels can enhance estimation accuracy by accounting for natural light/dark asymmetries. Furthermore, a diversity of inputs to motion detecting neurons can provide access to more complex higher-order correlations. Collectively, these results illustrate how non-canonical computations improve motion estimation with naturalistic inputs. This argues that the complexity of the fly's motion computations, implemented in its elaborate circuits, represents a valuable feature of its visual motion estimator.

Show MeSH