Discriminative model

As explained in Graph Neural Networks in Action

A discriminative model learns to distinguish among data classes or predict target values without modeling the full data-generating distribution.

A discriminative model learns to separate data classes or predict target values from input data. It focuses on decision boundaries rather than modeling the full data-generating distribution. section 5.0 section 5.1.1

A typical supervised classification workflow is:

• Train on labeled examples, treating the input as X and the desired target as Y. section 5.1 section 5.1.1
• Learn the conditional relationship P(Y|X), such as whether an image is a boat or an airplane. section 5.1 section 5.1.1
• Apply the learned boundary to unseen test data and return a predicted target label. section 5.1.1

Convolutional and attention-based graph neural networks are examples of discriminative models. They can identify patterns for graph-learning tasks such as node classification without learning the entire graph data distribution. section 5.0

Boat-versus-airplane classification

A model may distinguish boats from airplanes using telltale features such as a mast, keel, or boom. It need not model broader correlations such as airplanes having wings or boats appearing near water. section 5.1.1

Figure 5.1 Comparison of generative and discriminative tasks. On the left, the discriminative model learns to separate different images of boats and airplanes. On the right, the generative model attempts to learn the entire dataspace, which allows for new synthetic examples to be created such as a boat in the sky or an airplane on water.

Compare discriminative boundary separation with generative modeling of the entire dataspace and synthetic-example creation. section 5.1.1

How Graph Neural Networks in Action covers it

How deep it goes

Defines conditional prediction and boundary learning, stopping short of model internals or detailed training procedures.

Examples

Uses boat-versus-airplane images, telltale visual features, and node-classification tasks as examples.

What it assumes

Assumes labeled training examples and targets while defining training data, test data, and conditional probability from those elements.

What you can do after

Evaluate when boundary-focused classification is preferable to full-distribution modeling, considering computational cost and outlier robustness.

How it connects

Connects supervised learning, conditional probability, graph neural networks, generative models, and node classification.

liveBookmanning.com

Read Graph Neural Networks in Action on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Data Without LabelsA discriminative model predicts or classifies outcomes by modeling distinctions between data classes or conditional outcomes rather than generating new data.
• 

  liveBookmanning.com
  Investing for ProgrammersA discriminative model predicts labels or quantitative outcomes from input data without generating new content.
• 

  liveBookmanning.com
  Learn Generative AI with PyTorchA model that learns to predict labels from inputs by estimating conditional probabilities and decision boundaries between classes.

Compare how 4 books explain discriminative model →

Related concepts

• generative modelingcontrasts with · Generative modeling is a machine-learning approach that learns patterns or a data distribution to generate new samples resembling the data.
• machine learningequivalent to · Machine learning is a field of computational methods in which systems learn patterns or relationships from data to make predictions, decisions, or other outputs.
• conditional probabilityuses · Conditional probability is the probability of an event or variable value given that specified information, an event, or another variable’s value is known.
• generationcontrasts with · Generation is the computational production of outputs from supplied inputs or context, often by applying a model to produce predictions or new content.
• decision boundaryuses · A decision boundary is the locus in input or feature space that separates regions assigned different classes or decision outcomes by a predictive model.
• supervised learningused by · Supervised learning is a machine-learning approach that learns to map inputs to target outputs from examples paired with known labels or values.

This explainer was generated by Manning's knowledge graph from the text of Graph Neural Networks in Action. The section links point to the parts of the book it draws on.