tdhook.latent.dimension_estimation.local_knn

Classes

LocalKnnDimensionEstimator

Local intrinsic dimension estimation via k-NN distances [20].

Functions

_resolve_k(k, n)	Resolve k to an integer. If 'auto', use int(n**0.5), clamped to valid range.
_local_knn(data, k, eps)	Compute per-point local dimension. data: (N, D). Returns (N,) dimension estimates.

Module Contents

tdhook.latent.dimension_estimation.local_knn._resolve_k(k, n)

Resolve k to an integer. If ‘auto’, use int(n**0.5), clamped to valid range.

Parameters:

• k (Union[int, Literal['auto']])

• n (int)

Return type:

int

class tdhook.latent.dimension_estimation.local_knn.LocalKnnDimensionEstimator(k='auto', in_key='data', out_key='dimension', eps=1e-05)

Bases: tensordict.nn.TensorDictModuleBase

Local intrinsic dimension estimation via k-NN distances [20].

For each point x, d(x) = ln(2) / ln(R2k/Rk), where Rk and R2k are distances to the k-th and 2k-th nearest neighbors respectively.

Reads a data tensor from the input TensorDict. Expects (N, D) or (…, N, D). Outputs per-point dimension estimates of shape (…, N).

Parameters:

• k (Union[int, Literal['auto']])

• in_key (str)

• out_key (str)

• eps (float)

k = 'auto'

in_key = 'data'

out_key = 'dimension'

eps = 1e-05

in_keys

out_keys

forward(td)

Parameters:

td (tensordict.TensorDict)

Return type:

tensordict.TensorDict

__repr__()

tdhook.latent.dimension_estimation.local_knn._local_knn(data, k, eps)

Compute per-point local dimension. data: (N, D). Returns (N,) dimension estimates.

Parameters:

• data (torch.Tensor)

• k (int)

• eps (float)

Return type:

torch.Tensor