Feature matrix
A feature matrix is a rectangular array whose entries represent feature values associated with a collection of observations, entities, or other modeled objects.
Graph Neural Networks in Action
A matrix whose entries encode feature values for a collection of entities or graph elements.
- How it is used
- Represents graph-element attributes that generative graph models reconstruct alongside connectivity.
- How deep it goes
- Defines the matrix’s reconstruction role and implementation outputs, but does not detail feature encoding semantics.
- Examples
- Uses molecular node features in `data.x`, decoded feature matrices, and generated small-molecule graphs.
- What you can do after
- Build it from scratch using PyTorch decoder layers that emit feature and adjacency matrices for graph reconstruction.
Read how Graph Neural Networks in Action explains feature matrix →
Machine Learning for Tabular Data: XGBoost, Deep Learning, and AI
A rectangular numerical data structure in which rows represent observations and columns represent input features.
- How it is used
- Organizes numeric observations and input features for linear regression prediction.
- How deep it goes
- Explains numeric conversion, coefficient multiplication, missing values, outliers, and feature-target relationships, but not matrix algorithms internally.
- Examples
- Uses one-hot encoded categorical columns, numeric features, a coefficient vector, and a preprocessing pipeline.
- What you can do after
- Apply it by preparing numeric feature columns for regression and checking data quality before training.
Read how Machine Learning for Tabular Data: XGBoost, Deep Learning, and AI explains feature matrix →
Related concepts
- graph convolutional networkused by · A graph convolutional network is a graph neural network that computes representations by transforming and aggregating information from graph neighborhoods.
- decoderproduced by · A decoder is a component that maps encoded data or representations to decoded values, reconstructed data, or generated output.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.