What machine learning/deep learning technique to use for a data set consisting of 2D plane

Question

I'm new to machine learning/deep learning field. In an assignment I've been given a data set of 1000 data points. Each data point is a 2D place of dimension 1024 x 1024. In each plane there are some points marked as class-0 (x) and some points marked as class-1 (o). A simple diagram below.

data point #1

        (0,1024)                          (1024,1024)
            |--------------------------------|
            |          x                     |
            |        xx                      |
            |         x                      |
            |        x    o                  |
            |              x                 |
            |         x      x               |
            |                  o             |
            |                                |
            |                                |
            |                                |
            ----------------------------------
           (0,0)                           (1024,0)

data point #2

        (0,1024)                          (1024,1024)
            |--------------------------------|
            |                                |
            |        xo                      |
            |         o                      |
            |        x                       |
            |                                |
            |         x                      |
            |                  o             |
            |                   xxx          |
            |                                |
            |                                |
            ----------------------------------
           (0,0)                           (1024,0)

etc.

We know the Cartesian coordinate (x,y) of each labeled point.

What ML/DL technique should I use to make prediction for unlabeled data point like this.

unknown data point

        (0,1024)                          (1024,1024)
            |--------------------------------|
            |          .                     |
            |        ..                      |
            |         .                      |
            |        .                       |
            |       .                        |
            |      .                         |
            |                  .             |
            |                                |
            |                ..              |
            |               .                |
            ----------------------------------
           (0,0)                           (1024,0)

Any help is appreciated.

What attributes of the input would allow you to determine whether a sample is class 1 or class 0? Distance from some marker? Vertical or horizontal position? Something else? Do you know that there is a relationship between the information you have and class membership? — Sycorax, Jun 01 '19 at 21:22

score 0 · Answer 1 · answered Jun 01 '19 at 21:01

It sounds like you're asking about coming up with a good representation for the input.

You could rasterize these point sets as you've done above and use a 2D convolutional architecture. Alternatively you could use a continuous convolution model or a PointNets-like architecture on the raw point sets.

What machine learning/deep learning technique to use for a data set consisting of 2D plane

1 Answers1