MT population

The <pop> population object is a generic container for a population of units (model neurons) that are organized on a Cartesian x,y,z grid. Some of the major parameters and objects within the MT population are described here, whereas others are described in pages that are linked below.

  <pop>
    name mt
    area mt

    <geometry>
      xn   4           # population width (units)
      yn   4           # population height (units)
    </geometry>

    <mt_stf>
      <weight_config>
        type         QOC_2016        # Weight configuration:
                                     #   QOC_2016 - Quaia, Optican, Cumming (2016)
        velocity_x     8.0  (deg/s)  # Preferred velocity, x-component
        velocity_y     0.0  (deg/s)  # Preferred velocity, y-component
        dist_sd_pref   1.0           # SD for Gaussian distance weight, preferred
        dist_sd_anti   1.0           # SD for Gaussian distance weight, opposite
        low_tf_sd      1.1  (cyc/s)  # SD for Gaussian suppression of low TFs
        low_tf_k       0.5           # Scale factor for excitatory low TF suppr.
        write_w_table  zz.w.list     # Write a table of weights for STF channels
      </weight_config>

      negative_w_scale  0.0          # Scale negative weights by this factor

      <nonlin>
        type           half-rect     # Type of nonlinearity, "half-rect" ...
      </nonlin>
    </mt_stf>

    <input>   # Input from V1
      ...  See MT input
    </input>

    <spike_gen>   # Spike generation
      ...  See MT spike generation
    </spike_gen>
  </pop>

The name parameter can be chosen arbitrarily and is used to refer to this population.

The area parameter specifies a frame of reference for this population. In this case, it is the MT area previously described.

The <geometry> object specifies the size of the population. In this case, it is a 4 x 4 spatial grid. Because zn is not specified, the third (z) dimension is assumed to have depth of 1. Values for parameters not specified in the geometry may be inherited from the <area> object.

The <input> object configures the spatial convergence from V1 to MT and is described elsewhere.

The <spike_gen> object controls MT spike generation and is described elsewhere.

The <mt_stf> object contains parameters that govern how signals from the STF and direction channels are combined and transformed to generate the time-varying MT response.

The <weight_config> object contains parameters that specify the connection weighting scheme from the V1 STF channels to the MT unit.

• type - specifies the type of overall weighting scheme. Possible values are:
  • QOC_2016 - The weights will be determined using the methods described in Quaia, Optican and Cumming (2016, pages 3906-3907). The following parameters must be specified:
    • velocity_x - x-velocity of excitatory plane
    • velocity_y - y-velocity of excitatory plane
    • dist_sd_pref - SD of Gaussian to weight distance from the preferred velocity plane
    • dist_sd_anti - SD of Gaussian to weight distance from the anti-preferred velocity plane (opposite to that of the preferred plane).
    • low_tf_sd - SD of Gaussian to apply suppression to weights for filters that have low TFs (see sigma in Eqn 1 of Quaia et al., page 3907).
    • low_tf_k - amplitude scaling constant for the suppression of low TF excitatory weights (see K in Eqn 1 of Quaia et al., page 3907).
    • write_w_table - specifies a file name for writing a text table that lists the weights for each STF and direction channel.

      Example table format.

  • FILE - The weights will be read from a text file. [Not implemented yet.]

The negative_w_scale parameter allows negative V1-to-MT wieghts to be scaled by the specified value. This is useful for suppressing negative weights to create behavior that is more Component-cell like.

The <nonlin> object specifies the type of non-linearity apply to the MT output.