clustering-tfjs/dist/utils/tensor.d.ts

High-performance TypeScript clustering algorithms (K-Means, Spectral, Agglomerative) with TensorFlow.js acceleration and scikit-learn compatibility

import * as tf from '../tf-adapter';
/**
 * Converts a regular (nested) JavaScript array into a TensorFlow.js tensor
 * with the provided dtype (defaults to `float32`).
 *
 * The function is wrapped in `tf.tidy` to ensure that any intermediate
 * tensors that may be created by TensorFlow.js during conversion are
 * automatically disposed of.
 */
export declare function arrayToTensor(arr: tf.TensorLike, dtype?: tf.DataType): tf.Tensor;
/**
 * Converts a tensor back to a JavaScript array (synchronously).
 *
 * The returned value is a *copy* of the underlying data, so further
 * manipulations will not affect the original tensor.
 */
export declare function tensorToArray(tensor: tf.Tensor): number[] | number[][] | number[][][];
/**
 * Computes the element-wise Euclidean (ℓ2) distance between two tensors along
 * their last dimension.
 *
 * Both inputs must be broadcast-compatible. The result will have the broadcast
 * shape of `tf.sub(a, b).sum(-1)` (i.e. the shapes minus the last dimension).
 *
 * Example:
 * ```ts
 * const a = tf.tensor([[0, 0], [1, 1]]); // (2, 2)
 * const b = tf.tensor([1, 0]);           // (2)
 * euclideanDistance(a, b)  // => Tensor([1, 1])
 * ```
 */
export declare function euclideanDistance(a: tf.Tensor, b: tf.Tensor): tf.Tensor;
/**
 * Computes the Manhattan (ℓ1) distance between two tensors along their last
 * dimension.
 */
export declare function manhattanDistance(a: tf.Tensor, b: tf.Tensor): tf.Tensor;
/**
 * Computes the cosine distance (1 ‑ cosine similarity) between two tensors
 * along their last dimension.
 */
export declare function cosineDistance(a: tf.Tensor, b: tf.Tensor): tf.Tensor;
/**
 * Efficiently computes pairwise Euclidean distance matrix for a set of points
 * represented by a 2-D tensor of shape `(n, d)`.
 *
 * The implementation uses the well-known trick
 * ‖x − y‖² = ‖x‖² + ‖y‖² − 2·xᵀy with broadcasting to avoid allocating an
 * `(n, n, d)` intermediate tensor.
 */
export { pairwiseEuclideanMatrix } from './pairwise_distance';
//# sourceMappingURL=tensor.d.ts.map