Sparse coding models demonstrate some non-classical receptive field effects

Non-classical receptive field (nCRF) effects include several response properties in V1 neurons not explained by a linear-nonlinear (LN) receptive field model, but instead requiring significant interactions between V1 neurons. Using a sparse coding model [1,2] and bar and grating stimuli, simulated physiology experiments were carried out that replicated several nCRF phenomena reported previously in neurophysiology experiments. These include: end-stopping [3] (Fig. 1), contrast invariance of orientation tuning [4] (Fig. 2), radius, orientation, and contrast tunings of surround suppression [5,6] (Fig. 3, 4, 5). The results suggest that a sparse coding model can explain many of the nonlinear effects in V1 cells, and is therefore a reasonable candidate for a functional model of striate cortex.

Figure 1.

End-stopping. Comparison with a LN model.

Figure 2.

Contrast invariance of orientation tuning.

Figure 3.

Surround suppression at different contrasts.

Figure 4.

Orientation tuning of surround suppression.

Figure 5.

Surround orientation influences contrast tuning.