hash_table

torchhd.hash_table(keys: VSATensor, values: VSATensor) → VSATensor[source]

Hash table from keys-values hypervector pairs.

\[\bigoplus_{i = 0}^{n - 1} K_i \otimes V_i\]

Parameters:

keys (VSATensor) – The keys hypervectors, must be the same shape as values.
values (VSATensor) – The values hypervectors, must be the same shape as keys.

Shapes:

Keys: \((*, n, d)\)
Values: \((*, n, d)\)
Output: \((*, d)\)

Examples:

>>> k = torchhd.random(2, 6)
>>> k
tensor([[-1., -1., -1.,  1.,  1.,  1.],
        [-1.,  1.,  1., -1., -1.,  1.]])
>>> v = torchhd.random(2, 6)
>>> v
tensor([[-1.,  1.,  1.,  1., -1., -1.],
        [-1., -1.,  1., -1., -1., -1.]])
>>> torchhd.hash_table(k, v)
tensor([ 2., -2.,  0.,  2.,  0., -2.])