Skip to content

Selfsupervised

continuiti.data.selfsupervised

Self-supervised data set.

SelfSupervisedOperatorDataset(x, u)

Bases: OperatorDataset

A SelfSupervisedOperatorDataset is a data set that contains data for self-supervised learning. Every data point is created by taking one sensor as a label.

Every observation consists of tuples (x, u, y, v), where x contains the sensor positions, u the sensor values, and y = x_i and v = u_i are the label's coordinate its value for all i.

PARAMETER DESCRIPTION
x

Sensor positions of shape (num_observations, coordinate_dim, num_sensors...)

TYPE: Tensor

u

Sensor values of shape (num_observations, num_channels, num_sensors...)

TYPE: Tensor

Source code in src/continuiti/data/selfsupervised.py 
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
def __init__(self, x: torch.Tensor, u: torch.Tensor):
    self.num_observations = u.shape[0]
    self.coordinate_dim = x.shape[1]
    self.num_channels = u.shape[1]

    # Check consistency across observations
    for i in range(self.num_observations):
        assert (
            x[i].shape[0] == self.coordinate_dim
        ), "Inconsistent coordinate dimension."
        assert (
            u[i].shape[0] == self.num_channels
        ), "Inconsistent number of channels."

    xs, us, ys, vs = [], [], [], []

    x_flat = x.view(self.num_observations, self.coordinate_dim, -1)
    u_flat = u.view(self.num_observations, self.num_channels, -1)
    self.num_sensors = math.prod(u.shape[2:])

    for i in range(self.num_observations):
        # Add one data point for every sensor
        for j in range(self.num_sensors):
            y = x_flat[i, :, j].unsqueeze(0)
            v = u_flat[i, :, j].unsqueeze(0)

            xs.append(x[i])
            us.append(u[i])
            ys.append(y)
            vs.append(v)

    xs = torch.stack(xs)
    us = torch.stack(us)
    ys = torch.stack(ys)
    vs = torch.stack(vs)

    super().__init__(xs, us, ys, vs)
Last update: 2024-08-20
Created: 2024-08-20