binoc_filter
WM Working Models www.iModel.org/wm
Overview
This model implements the basic binocular disparity energy (BDE) model of Ohzawa et al. (1998), as well as variations on this model proposed by Chen et al. (2001) and Read et al. (2002).
Configuring the Model
Parameters
These parameters set the basic simulation settings:The following phase parameters are common to all BDE models and set up the spatial phase shift between the left and right eyes. Note that in WM, the negative spatial filters are inferred (we do not have explicit parameters for setting the -1s/d or -2s/d pathways).
- sscale - spatial resolution (degrees/pixel)
- tscale - temporal resolution (samples per second)
- xn - filter x-axis spatial extent (pixels)
- yn - filter y-axis spatial extent (pixels)
- tn - duration of the model (sampling units)
The next 2 options are implemented for defining the 4 subunit filters in the BDE models. ME models are always in a specific quadrature relationship (+90 deg) by convention, but this can produce incorrect filters for the Read et al. models depending on which type of disparity tuning is being built (ie. TE/TI vs NE/FA).
- phase_1 - Spatial phase of cosine in Gabor for first left filter
- phase_shift - Spatial phase shift of right eye RF relative to phase_1
The following parameters relate to the last two stages of the BDE model, introducing options for rectification and inhibitory interactions as in the Read et al. models. In particular, 'right_sign' sets the pattern of +'s and -'s that indicate addition or subtraction of each subunit output.
- phase_shift_opp - [0,1] -- if 1, use negative phase shift to generate the RF phase for the second pathway (i.e., 2s = 1s - phase_shift)
- quad_shift_opp - [0,1] -- if 1, use -90 deg quadrature shift instead of +90 (2s to 2d). This is allows a variation on quadrature from the ME models, where the quadrature shift is always +90 degrees (ME model we define 1s to 1d shift as +90, 2s and 2d are then adjusted to give anti-direction RFs)
- simp_rect - [0,1] If 1, monocular filter outputs are rectified before summation
- simp_thresh - Non-zero threshold to be applied to monocular filter outputs (fraction of filter area)
- binoc_nonlin - type of nonlinearity for final stage [squaring, halfsq]
- right_sign - [+,-] add/subtract a given subunit. Symbols represent, in order, subunits 1s 1d 2s 2d
Filters
binoc_filter models use a <filter> object that specifies parameters for the spatiotemporal RFs of each monocular subunit. For the BDE_Gabor and BDE_Gabor_DS models as well as the BDE_RPC models, the basic 'Gabor' filter is used. With the BDE_CWQ models, a specific 'Gabor_CWQ' filter type is defined, that includes parameters for the temporal filtering implented in the Chen et al. models.
Spike Generation
binoc_filter models also require a <spike_gen> object that defines the method of spike generation from the final filter output.
Model Outputs
Model Components
The spatial filters for each of the four component monocular subunits can be output using the <filter> object's write_filter option. If Poisson spike generation is used, then the output firing probability can be written to a .pl file using the spike_dump option in <spike_gen>.Model Responses
The following response types can be requested:A template .rsp file including these quantities can be downloaded here.
- spikes - spike times
- binoc_energy - final stage filter output
- binoc_1 - input to final stage from pathway 1
- binoc_2 - input to final stage from pathway 2
- filter_left_1 - output from subunit 1s
- filter_left_2 - output from subunit 2s
- filter_right_1 - output from subunit 1d
- filter_right_2 - output from subunit 2d
- binoc_1_pos - summed output from 1s and 1d
- binoc_1_neg - summed output from -1s and -1d
- binoc_2_pos - summed output from 2s and 2d
- binoc_2_neg - summed output from -2s and -2d
- binoc_1_pos_sq - binoc_1_pos signal squared
- binoc_1_neg_sq - binoc_1_neg signal squared
- binoc_2_pos_sq - binoc_2_pos signal squared
- binoc_2_neg_sq - binoc_2_neg signal squared
Models of This Type