SparseHD

class torchhd.classifiers.SparseHD(n_features: int, n_dimensions: int, n_classes: int, *, n_levels: int = 100, min_level: int = -1, max_level: int = 1, epochs: int = 120, lr: float = 0.035, sparsity: float = 0.1, sparsity_type: Literal['dimension', 'class'] = 'dimension', device: device | None = None, dtype: dtype | None = None)[source]

Implements SparseHD: Algorithm-Hardware Co-optimization for Efficient High-Dimensional Computing.

Parameters:

  • n_features (int) – Size of each input sample.

  • n_dimensions (int) – The number of hidden dimensions to use.

  • n_classes (int) – The number of classes.

  • n_levels (int, optional) – The number of discretized levels for the level-hypervectors.

  • min_level (int, optional) – The lower-bound of the range represented by the level-hypervectors.

  • max_level (int, optional) – The upper-bound of the range represented by the level-hypervectors.

  • epochs (int, optional) – The number of iteration over the training data.

  • lr (float, optional) – The learning rate.

  • sparsity (float, optional) – The fraction of weights to be zero.

  • sparsity_type (str, optional) – The way in which to apply the sparsity, per hidden dimension, or per class.

  • device (torch.device, optional) – the desired device of the weights. Default: if None, uses the current device for the default tensor type (see torch.set_default_tensor_type()). device will be the CPU for CPU tensor types and the current CUDA device for CUDA tensor types.

  • dtype (torch.dtype, optional) – the desired data type of the weights. Default: if None, uses torch.get_default_dtype().

__call__(samples: Tensor) → Tensor

Evaluate the logits of the classifier for the given samples.

Parameters:

samples (Tensor) – Batch of samples to be classified.

Returns:

Logits of each sample for each class.

Return type:

Tensor

accuracy(data_loader: Iterable[Tuple[Tensor, LongTensor]]) → float

Accuracy in predicting the labels of the samples.

Parameters:

data_loader (DataLoader) – Iterable of tuples containing a batch of samples and labels.

Returns:

The accuracy of predicting the true labels.

Return type:

float

fit(data_loader: Iterable[Tuple[Tensor, LongTensor]])[source]

Fits the classifier to the provided data.

Parameters:

data_loader (DataLoader) – Iterable of tuples containing a batch of samples and labels.

Returns:

self

predict(samples: Tensor) → LongTensor

Predict the class of each given sample.

Parameters:

samples (Tensor) – Batch of samples to be classified.

Returns:

Index of the predicted class for each sample.

Return type:

LongTensor