Skip to content

Getting Started

Installation

Install the core package with pip:

pip install cca-zoo

The core package requires Python ≥ 3.10 and provides all linear and nonparametric methods.

Optional extras

Install extras for additional method families:

pip install cca-zoo[deep]          # DCCA variants (PyTorch + Lightning)
pip install cca-zoo[probabilistic] # Probabilistic CCA (NumPyro + JAX)
pip install cca-zoo[all]           # Everything above

PyTorch installation

For GPU support or specific CUDA versions, follow the PyTorch installation guide before installing cca-zoo[deep].

Core concepts

Views

CCA-Zoo expects data as a list of arrays, one per view:

views = [X1, X2]          # two views
views = [X1, X2, X3]      # three views

Each array has shape (n_samples, n_features_i). All views must share the same number of rows.

The estimator API

Every model follows the same three-step pattern:

from cca_zoo.linear import CCA

model = CCA(latent_dimensions=2)  # 1. construct
model.fit(views)                   # 2. fit
z = model.transform(views)         # 3. use

Or equivalently:

z = CCA(latent_dimensions=2).fit_transform(views)

Evaluating fit quality

score returns the average pairwise canonical correlation per latent dimension:

corrs = model.score(views)   # np.ndarray, shape (latent_dimensions,)
print(corrs)                  # e.g. [0.94, 0.87]

Inspecting weights

After fitting, model.weights is a list of weight matrices (one per view):

W1, W2 = model.weights   # each shape (n_features_i, latent_dimensions)

Quick start

Two-view CCA

import numpy as np
from cca_zoo.datasets import JointData
from cca_zoo.linear import CCA

# Simulate data from a linear latent variable model
data = JointData(
    n_views=2,
    n_samples=200,
    n_features=[50, 50],
    latent_dimensions=2,
    signal_to_noise=2.0,
    random_state=0,
)
train_views = data.sample()
test_views  = data.sample()

# Fit and evaluate
model = CCA(latent_dimensions=2).fit(train_views)
print("Canonical correlations:", model.score(test_views))

# Project into the shared latent space
z1, z2 = model.transform(test_views)
print("Latent shape:", z1.shape)  # (200, 2)

Multiview CCA (≥2 views)

from cca_zoo.linear import MCCA

data = JointData(n_views=3, n_samples=200, n_features=30, random_state=0)
views = data.sample()

model = MCCA(latent_dimensions=2).fit(views)
print(model.score(views))

Regularised CCA

from cca_zoo.linear import rCCA

# c controls the ridge penalty (0 = CCA, 1 = PLS)
model = rCCA(latent_dimensions=2, c=0.1).fit(train_views)

Kernel CCA

from cca_zoo.nonparametric import KCCA

model = KCCA(latent_dimensions=2, kernel="rbf", gamma=0.01, c=0.1).fit(train_views)
z1, z2 = model.transform(test_views)

CCA-Zoo's GridSearchCV wraps sklearn's grid search with a multiview interface:

from cca_zoo.model_selection import GridSearchCV
from cca_zoo.nonparametric import KCCA

param_grid = {"c": [0.01, 0.1, 1.0], "gamma": [0.01, 0.1]}
gs = GridSearchCV(KCCA(latent_dimensions=2, kernel="rbf"), param_grid, cv=5)
gs.fit(train_views)
print("Best params:", gs.best_params_)

Next steps