Dayan & Abbott Meeting 4 (Chapter 2, 2.4 onwards)

Recap Nyquist Frequency

A nice example of temporal aliasing in case people missed it.

Fig. 2.9 example.

Separable Receptive Fields

  1. Measuring neuron response to images by three variables $x, y$ locations and $\tau$ time.
  2. Some neurons have separable or non-separable receptive fields. $D(x,y,\tau) = D_S(x,y)D_t(\tau)$ (separate spatial and temporal).

Spatial Receptive Fields

  1. Figure 2.10 x/y axes are spatial location. z axis is $(D(x,y,\tau))$, which is how strongly the visual stimulus affects the firing rate.
  2. Separate ON and OFF regions.
  3. Figure 2.11. Can see expected response by superimposing receptive field over stimuli. Shows the benefit of OFF part of field.
  4. Figures show cells with 2 subregions, but cells typically 1-5.

The following code plots equation 2.27. However, I've added the term $sin(ky-\theta)$ to the end of the equation. I'm not sure if this is correct (but without it I don't see how the equation is meant to lead to the receptive fields as shown in the figures). It's probable that a different change is correct, or that I just made a mistake in the code.

Temporal Receptive Fields

  1. Figure 2.13 shows magnitude and sign varying, but location remaining fairly constant (approximate separability).
  2. Figure 2.17 plots against $\tau$ showing variation over time.