MT input

The <input> object is used to specify the spatial distribution of the V1 to MT inputs. The parameters are described below for the following example:

  <input>
    name        v1
    type        plain       # not a 'pop' type of input
    dir_config  stacked     # ['stacked'] or 'random'
    pop_origin  v1          # Population of origin of inputs

    <distrib>                 # Spatial distribution of V1 inputs
      type     distance       # Connection probability depends on distance
      cdist       0.6  (deg)  # (deg) crit connection distance (SD)
      shape    Gaussian       # Shape of distribution
      minw        0.05        # () suppress connections with lower weight
      seed     1777           # () Randomization seed for selecting inputs
      nsamp      60           # () number of inputs to select; 2-special geom.
      even_out    1           # 1-even out the distribution; 0-do not
    </distrib>
  </input>

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

The type parameter is used to indicate that this is a generic type of input, i.e., a plain input.

The dir_config parameter specifies how direction preference and spatial location are varied. A value of stacked means that for each spatial location that provides input to the MT cell, all direction channels at that location, and within that particular STF channel, will connect to the MT cell. A value of random indicates that direction and location vary randomly and independently.

The pop_origin parameter indicates that the inputs will come from v1. Note, the v1 population is not explicitly defined in the model parameter file, but is implicitly configured as described here. This parameter currently serves no function in the model simulation, but is kept here for consistency and documentation.

The <distrib> object specifies the spatial distribution of the the V1 inputs. Its parameters are as follows:

• type - this specifies distance, indicating that the connection probability will be determined by the distance of the V1 unit RF center from the MT unit RF center.

• cdist - specifies the critical distance parameter, here 0.6 deg of visual angle. Note, this distance relates to the stimulus grid resolution sscale.

• shape - specifies that the probability distribution is Gaussian, and thus, cdist (above) will be the SD of that Gaussian.

• minw - stipulates that no connection will be made where the Gaussian (which has a value of 1 at its center) falls below this value (0.05 here).

• seed - the randomization seed for picking the random spatial locations for inputs.

• nsamp - the number of V1 inputs to select, here 60.

• even_out - Whether to keep the first 60 unique input locations chosen (0), or to even out (1) the inputs so that the V1 inputs have more even coverage.

The following figure shows an example of the spatial connectivity achieved from the 32 x 32 unit V1 population (right) to one of the units in the 4 x 4 MT population (left, white box marks the particular MT unit). In this case, the marked MT unit will integrate the signals from the 60 locations marked, from each of 12 direction channels (not shown) at each location. This synaptic input map shows inputs for one of the STF channels. Inputs from the other STF channels will have independently chosen spatial locations.

Inputs for other MT units will be independently chosen, and will be shifted in the 32x32 map so that they remain centered on the central coordinate of that particular MT unit (not shown). The correspondence between MT center locations and the V1 / stimulus map is given here.