Skip to main content
Version: v1.5.0

Supported problem types

Overview​

H2O Hydrogen Torch supports an array of diverse problem types in computer vision, natural language, and audio, and each, in turn, is explained below.

Image​

Image regression​

  • Description: Image regression refers to assigning one or more continuous target labels to an input image.
  • Supported regression tasks: Multi-label.
note

To learn about the available settings for this problem type, see Experiment settings: Image regresssion.

3D image regression​

  • Description: 3D image regression refers to assigning one or more continuous target labels to 3D images (inputs).
  • Supported regression tasks: Multi-label.
note

To learn about the available settings for this problem type, see see Experiment settings: 3D image regresssion.

Image classification​

  • Description: Image classification refers to assigning one or more categorical target labels to an input image.
  • Supported classification tasks: Binary, multi-class, and multi-label.
note

To learn about the available settings for this problem type, see Experiment settings: Image classification.

3D image classification​

  • Description: 3D image classification refers to assigning one or more categorical target labels to 3D images (inputs).
  • Supported classification tasks: Binary, multi-class, and multi-label.
note

To learn about the available settings for this problem type, see Experiment settings: 3D image classification.

Image object detection​

  • Description: Image object detection refers to locating individual objects in an image by drawing bounding boxes around them.
  • Supported classification tasks: Multi-class.
note

To learn about the available settings for this problem type, see Experiment settings: Image object detection.

Image semantic segmentation​

  • Description: Image semantic segmentation refers to associating each pixel of an image with a class label (such as phones, pencils, or roads).
  • Supported classification tasks: Multi-class.
note

To learn about the available settings for this problem type, see Experiment settings: Image semantic segmentation.

3D image semantic segmentation​

  • Description: 3D image semantic segmentation refers to associating each pixel of a 3D image with a particular class label.
  • Supported classification tasks: Multi-class.
note

To learn about the available settings for this problem type, see Experiment settings: 3D image semantic segmentation.

Image instance segmentation​

  • Description: Image instance segmentation refers to locating individual objects in an image by drawing masks around them.
  • Supported classification tasks: Multi-class.
note

To learn about the available settings for this problem type, see Experiment settings: Image instance segmentation.

Image metric learning​

  • Description: Image metric learning refers to establishing similarity or dissimilarity between images.
note

To learn about the available settings for this problem type, see Experiment settings: Image metric learning.

Text​

Text regression​

  • Description: Text regression refers to assigning one or more continuous target labels to an input text.
  • Supported regression tasks: Multi-label.
note

To learn about the available settings for this problem type, see Experiment settings: Text regression.

Text classification​

  • Description: Text classification refers to assigning one or more categorical target labels to an input text; this includes binary classification, multi-class classification, and multi-label classification.
  • Supported classification tasks: Binary, multi-class, and multi-label.
note

To learn about the available settings for this problem type, see Experiment settings: Text classification.

Text token classification​

  • Description: Text token classification refers to assigning a label to all tokens in a piece of text in contrast to text classification, where the entire text is given a label.
  • supported classification tasks: Binary and multi-class.
note

To learn about the available settings for this problem type, see Experiment settings: Text token classification.

Text span prediction​

  • Description: Text span prediction refers to the task of finding a substring in a larger context. A typical problem (task) is question-answering, where given a context and question, the task is to find the answer (substring) in the context.
note

To learn about the available settings for this problem type, see Experiment settings: Text span prediction.

Text sequence-to-sequence​

  • Description: Text sequence-to-sequence (Seq2seq) refers to the task of predicting an output sequence given an input sequence; in other words, Seq2seq turns one sequence into another (sequence transformation).
note

To learn about the available settings for this problem type, see Experiment settings: Text sequence-to-sequence.

Text metric learning​

  • Description: Text metric learning refers to establishing similarity or dissimilarity between texts.
note

To learn about the available settings for this problem type, see Experiment settings: Text metric learning.

Image and text​

Image and text classification​

  • Description: Image and text classification refers to assigning one or more categorical target labels to a combined input of image and text.
  • Supported classification tasks: Binary, multi-class, and multi-label.
note

To learn about the available settings for this problem type, see Experiment settings: Image and text classification.

Audio​

Audio regression​

  • Description: Audio regression refers to assigning one or more continuous target labels to input audio.
  • Supported regression tasks: Multi-label.
note

To learn about the available settings for this problem type, see Experiment settings: Audio regression.

Audio classification​

  • Description: Audio classification refers to assigning one or more categorical target labels to input audio.
  • Supported classification tasks: Binary, multi-class, and multi-label.
note

To learn about the available settings for this problem type, see Experiment settings: Audio classification.

Speech​

Speech recognition​

  • Description: Speech recognition refers to the process of converting input audio to text.
note

To learn about the available settings for this problem type, see Experiment settings: Speech recognition.

Graph​

Graph node classification​

  • Description: Graph node classification refers to assigning one or more categorical target labels to a specific node within a graph. This type of problem involves analyzing the relationships and connections between nodes in a graph structure to predict the appropriate label for a given node. The goal is to accurately classify each node based on its features and the characteristics of the surrounding nodes in the graph.
  • Supported classification tasks: Binary, multi-class, and multi-label.
note

To learn about the available settings for this problem type, see Experiment settings: Graph node classification.

Graph node regression​

  • Description: Graph node regression refers to assigning continuous target labels to one or more nodes within a graph. In particular, it refers to predicting one or more continuous target values for a specific node within a graph. This problem involves analyzing the relationships and connections between nodes in a graph structure to forecast the appropriate numerical value for a given node. The goal is to accurately predict the continuous target label for each node based on its features and the characteristics of the surrounding nodes in the graph.
  • Supported regression tasks: Multi-label.
note

To learn about the available settings for this problem type, see Experiment settings: Graph node regresssion.

Multi-modal​

Multi-modal causal language modeling​

  • Description: Multi-modal causal language modeling involves training or fine-tuning machine learning models to generate textual responses by leveraging both text-based queries and visual inputs, such as images.
note

To learn about the available settings for this problem type, see Experiment settings: Multi-modal causal language modeling.

Feedback