TheAlgorithms/Python · error · ValueError

n_components and n_iter must be >= 1

Error message

n_components and n_iter must be >= 1

What it means

Raised by apply_tsne in t_stochastic_neighbour_embedding.py when n_components or n_iter is less than 1. The t-SNE implementation initializes a random embedding of shape (n_samples, n_components) and iterates n_iter times, so zero or negative values for either parameter make the gradient-descent loop meaningless and are rejected up front.

Source

Thrown at machine_learning/t_stochastic_neighbour_embedding.py:114

    """
    Apply t-SNE for dimensionality reduction.

    Args:
        data_matrix: Original dataset (features).
        n_components: Target dimension (2D or 3D).
        learning_rate: Step size for gradient descent.
        n_iter: Number of iterations.

    Returns:
        ndarray: Low-dimensional embedding of the data.

    >>> features, _ = collect_dataset()
    >>> embedding = apply_tsne(features, n_components=2, n_iter=50)
    >>> embedding.shape
    (150, 2)
    """
    if n_components < 1 or n_iter < 1:
        raise ValueError("n_components and n_iter must be >= 1")

    n_samples = data_matrix.shape[0]
    rng = np.random.default_rng()
    embedding = rng.standard_normal((n_samples, n_components)) * 1e-4

    high_dim_affinities = compute_pairwise_affinities(data_matrix)
    high_dim_affinities = np.maximum(high_dim_affinities, 1e-12)

    embedding_increment = np.zeros_like(embedding)
    momentum = 0.5

    for iteration in range(n_iter):
        low_dim_affinities, numerator_matrix = compute_low_dim_affinities(embedding)
        low_dim_affinities = np.maximum(low_dim_affinities, 1e-12)

        affinity_diff = high_dim_affinities - low_dim_affinities

        gradient = 4 * (

View on GitHub (pinned to f5988cc097)

Solutions

1. Pass n_components >= 1 (2 or 3 are the usual choices for visualization).
2. Pass n_iter >= 1; typical values are 250-1000 for this implementation.
3. Guard computed parameter values before the call: max(1, requested) or explicit validation.

Example fix

# before
embedding = apply_tsne(features, n_components=0, n_iter=50)

# after
embedding = apply_tsne(features, n_components=2, n_iter=50)

n_components = max(1, int(n_components))
n_iter = max(1, int(n_iter))
embedding = apply_tsne(data_matrix, n_components=n_components, n_iter=n_iter)

When it happens

Trigger: Calling apply_tsne(data_matrix, n_components=0) or with n_iter=0 (or negative values for either), often from a config where an unset parameter defaults to 0.

Common situations: CLI/config-driven dimensionality reduction where n_components is computed as len(selected_columns)-something and underflows to 0, or a hyperparameter sweep boundary that includes 0.

Related errors

• x and y have different lengths
• Input arrays must have the same length.
• Test samples' feature length does not equal to that of train
• Expected a_coeffs to have {self.order + 1} elements for {sel
• Expected b_coeffs to have {self.order + 1} elements for {sel

AI-assisted analysis of TheAlgorithms/Python@f5988cc097 (2026-08-14). Data as JSON: /api/errors/b5ac8127dac456d1. Report an issue: GitHub.