Embedders are dimensionality reducers that produce dense feature representations from the samples contained in a dataset. They are a type of Transformer but they operate more like a Learner under the hood. In fact, the type of learning that Embedders employ is called manifold learning because it aims to find a low-dimensional manifold of a high-dimensional dataset. Embedders are typically used for visualizing high-dimensional datasets and for producing inputs to a learning algorithm.
Embed a Dataset#
To embed a dataset you can use the
apply() method on a dataset object like you would a regular Transformer.
use Rubix\ML\Datasets\Unlabeled; use Rubix\ML\Embedders\TSNE; $dataset = new Unlabeled($samples); echo $dataset->numColumns(); // 100 dimensions $dataset->apply(new TSNE(3, 200.0)); echo $dataset->numColumns(); // 3 dimensions