Truncated SVD#
Truncated Singular Value Decomposition (SVD) is a matrix factorization and dimensionality reduction technique that generalizes eigendecomposition to general matrices. When applied to datasets of document term frequency vectors, the technique is called Latent Semantic Analysis (LSA) and computes a statistical model of relationships between words.
Note
Note that the Tensor extension is required to use this transformer.
Interfaces: Transformer, Stateful, Persistable
Data Type Compatibility: Continuous only
Parameters#
# | Name | Default | Type | Description |
---|---|---|---|---|
1 | dimensions | int | The target number of dimensions to project onto. |
Example#
use Rubix\ML\Transformers\TruncatedSVD;
$transformer = new TruncatedSVD(100);
Additional Methods#
Return the proportion of information lost due to the transformation:
public lossiness() : ?float
References#
-
S. Deerwater et al. (1990). Indexing by Latent Semantic Analysis. ↩
Last update:
2021-03-07