bundle_sequence

torchhd.bundle_sequence(input: VSATensor) → VSATensor[source]

Bundling-based sequence.

The first value is permuted \(n-1\) times, the last value is not permuted.

\[\bigoplus_{i=0}^{n-1} \Pi^{n - i - 1}(V_i)\]

Parameters:: input (VSATensor) – The hypervector values.

Shapes:

Input: \((*, n, d)\)
Output: \((*, d)\)

Examples:

>>> x = torchhd.random(4, 6)
>>> x
tensor([[ 1., -1.,  1.,  1.,  1.,  1.],
        [-1.,  1., -1., -1.,  1., -1.],
        [ 1.,  1., -1., -1., -1.,  1.],
        [-1., -1.,  1., -1.,  1.,  1.]])
>>> torchhd.bundle_sequence(x)
tensor([ 2.,  0.,  2.,  0., -2.,  0.])