8000 TSNE is ignoring n_iter_without_progress parameter · Issue #7678 · scikit-learn/scikit-learn · GitHub
[go: up one dir, main page]
Skip to content

TSNE is ignoring n_iter_without_progress parameter #7678

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Closed
gioelelm opened this issue Oct 15, 2016 · 1 comment
Closed

TSNE is ignoring n_iter_without_progress parameter #7678

gioelelm opened this issue Oct 15, 2016 · 1 comment

Comments

@gioelelm
Copy link

Description

TSNE with method="exact" is ignoring the n_iter_without_progress parameter.
My insight is that it never gets passed to _gradient_descent:
opt_args['n_iter_without_progress'] is only set for barnes_hut and not for "exact"

Steps/Code to Reproduce

from sklearn.decomposition import PCA
from sklearn.manifold import TSNE
from sklearn.datasets import fetch_mldata
from numpy import *

# Load MNIST dataset
mnist = fetch_mldata("MNIST original")
X, y = mnist.data / 255.0, mnist.target

# Create subset and reduce to first 50 dimensions
indices = arange(X.shape[0])
random.shuffle(indices)
n_train_samples = 1000
X_pca = PCA(n_components=50).fit_transform(X)
X_train = X_pca[indices[:n_train_samples]]

X_train_embedded = TSNE(n_components=2, perplexity=40, verbose=2,\
method='exact', n_iter_without_progress=400, n_iter=1000).fit_transform(X_train)

Expected Results

[t-SNE] Iteration 225: did not make any progress during the last 400 episodes. Finished.

Actual Results

[t-SNE] Computing pairwise distances...
[t-SNE] Computed conditional probabilities for sample 1000 / 1000
[t-SNE] Mean sigma: 2.524818
[t-SNE] Iteration 25: error = 18.2748800, gradient norm = 0.0848871
[t-SNE] Iteration 50: error = 17.5069131, gradient norm = 0.0767555
[t-SNE] Iteration 75: error = 17.4644246, gradient norm = 0.0764557
[t-SNE] Iteration 100: error = 17.5864772, gradient norm = 0.0684978
[t-SNE] KL divergence after 100 iterations with early exaggeration: 17.586477
[t-SNE] Iteration 125: error = 1.0545128, gradient norm = 0.0067884
[t-SNE] Iteration 150: error = 1.0181383, gradient norm = 0.0080407
[t-SNE] Iteration 175: error = 1.0443270, gradient norm = 0.0086719
[t-SNE] Iteration 200: error = 1.0717048, gradient norm = 0.0089992
[t-SNE] Iteration 225: error = 1.0598099, gradient norm = 0.0094873
[t-SNE] Iteration 225: did not make any progress during the last 50 episodes. Finished.
[t-SNE] Error after 225 iterations: 1.059810

Versions

Darwin-15.5.0-x86_64-i386-64bit
('Python', '2.7.12 |Anaconda 4.2.0 (x86_64)| (default, Jul 2 2016, 17:43:17) \n[GCC 4.2.1 (Based on Apple Inc. build 5658) (LLVM build 2336.11.00)]')
('NumPy', '1.11.2')
('SciPy', '0.17.1')
('Scikit-Learn', '0.18')
@jnothman
Copy link
Member

Fixed in master (#6497). Will be released in 0.18.1

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
2 participants
@jnothman @gioelelm
0