The Model

Home
Background
Questions
The Model
References
Publications

I am a very visually oriented person.  I find that if a visual image can be made of an idea, I usually find it much easier to understand.  So, I want to build a model that will allow me visualize the output of the retina, beginning with the fovea.

Well, I have not really developed this model to any great extent, but some primitive ideas that will form the base of this model can be seen in the this powerpoint presentation.  Keep in mind, it is a module developed to introduce some of the implications of receptive fields to undergraduates.  Its summation model is primitive and only the equivalent of X cells are indicated.  One of the current great limitations of this demonstration is how fixed all the parameters of the model are. 

The current model development has proceeded along two lines: a simple one-dimensional model using excel, and a full-two dimensional model using Mathematica®.   The excel version is nice because its results are manipulable directly through the web on pages that I have posted, as long as you have the Office 2000tm web components on your computer.  You can begin these pages hereThe current limitations of this Excel model are briefly noted here.

The Mathematica® cannot be used directly through the web, but I am creating pages showing you the results and you can download the notebook files and run them on your machine if you so desire.   

Here is the directory where the files are located if you wish to download the Mathematica® files before I have the material organized.  

Here is the beginning of an organized presentation of the results of this mode.

Ultimately,  I would like to manipulate the following aspects of the model:

The image that is input
The size of the receptive fields
The strength of the lateral inhibition
The size of the central field of the receptive field
The amount of jitter in the image due to the movement of the eye
Whether the cells are X or Y in response type

This model also needs to be two-dimensional as well.

In the Mathematic version of the model you can:

  1.  input any grayscale, 8-bit input image or calculate your own image array,
  2. three sizes of receptive fields are output automatically in several version and the size of the field can be manipulated
  3. The output is two dimensional
  4. and I currently have the model adapt at the local level of the receptive field in the most advanced versions of the model.  I daresay the way adaptation happens is not very realistic but it is interesting to look at the responses anyway.