SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2606.04451 · 2026

On Out-of-sample Embedding in UMAP

Mohammad Islam, Jason Fleischer

arXiv · PDF · Open in the Atlas

Code that ran

1 of 1 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 1 failure of our instrument; 0 no recorded run), measured 2026-11-22; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2606.04451.json · badge: /api/ran/2606.04451.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
tariqul-islam/OOS_UMAP canonical (classifier label)
classifier basis: plan_row
1 ran · 0 no recorded run
FunctionStatusWhere it lives
load_mnist Ran
ran; our contract draft was wrong, not the code
tariqul-islam/OOS_UMAP/MNIST/mnist_reader.py
code served (permissive licence) · get_code("fcde486b88e624a8")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Sonnet 4.5, prompt python_fixture_v1) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import os
import gzip
import numpy as np
import tempfile
import shutil

# Create a temporary directory to hold mock MNIST files
temp_dir = tempfile.mkdtemp()

# Create mock MNIST data
num_samples = 10
image_size = 784  # 28x28

# Create mock training images and labels
mock_images = np.random.randint(0, 256, size=(num_samples, image_size), dtype=np.uint8)
mock_labels = np.random.randint(0, 10, size=num_samples, dtype=np.uint8)

# Write mock training images file
images_path = os.path.join(temp_dir, 'train-images-idx3-ubyte.gz')
with gzip.open(images_path, 'wb') as f:
    # MNIST image file format: magic number (4 bytes), num images (4 bytes), rows (4 bytes), cols (4 bytes), then data
    magic = np.array([0x00000803], dtype='>i4')  # Magic number for images
    num = np.array([num_samples], dtype='>i4')
    rows = np.array([28], dtype='>i4')
    cols = np.array([28], dtype='>i4')
    f.write(magic.tobytes())
    f.write(num.tobytes())
    f.write(rows.tobytes())
    f.write(cols.tobytes())
    f.write(mock_images.tobytes())

# Write mock training labels file
labels_path = os.path.join(temp_dir, 'train-labels-idx1-ubyte.gz')
with gzip.open(labels_path, 'wb') as f:
    # MNIST label file format: magic number (4 bytes), num labels (4 bytes), then data
    magic = np.array([0x00000801], dtype='>i4')  # Magic number for labels
    num = np.array([num_samples], dtype='>i4')
    f.write(magic.tobytes())
    f.write(num.tobytes())
    f.write(mock_labels.tobytes())

ARGS = {
    'path': temp_dir,
    'kind': 'train'
}
Setup block drafted by
a language model, Claude Sonnet 4.5, prompt python_fixture_v1
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block

Also present in other papers' code

Of 1 samples lifted from this paper's linked repositories, 1 carry a function identity and 0 do not. Of the 1 distinct functions with an identity, 1 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 0 the same month, 0 later) and 0 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

FunctionAlso present inEarliest-datedOthers named
load_mnist 2 other papers
2 dated earlier
1602.02697 (2016-02)
Practical Black-Box Attacks against Machine Learning
1708.07747 (2017-08)
Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Neighbor embedding algorithms reveal correlations in high-dimensional data by constructing an equivalent graph representation in a lower-dimensional space. An increasingly popular algorithm is Uniform Manifold Learning and Projection (UMAP), which uses algebraic topology to map distances between the two spaces. While it works well on many types of data sets, UMAP has trouble adding out-of-sample points to a pre-existing mapping. In particular, UMAP often places new points on the periphery of the found clusters, rather than in their interiors with their correlated neighbors. Here, we overcome this "repulsion effect" by optimizing pairwise interactions within the original k-nearest-neighbor graph. Moreover, we show that parameterizing UMAP obtains better embeddings than non-parametric algorithms, particularly as the data gets more complex (e.g., medical images). We also show that the repulsion effect is naturally mitigated when a parameterized UMAP is employed to embed the data. We characterize different UMAP approaches using trustworthiness, nearest neighbor classifiers, and by analyzing attractive and repulsive forces in the embeddings.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2606.04451")
get_code_for_paper("2606.04451")
have("2606.04451")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2606.04451.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2606.04451.svg)](https://syntology.ai/paper/2606.04451)

Connect an agent — have() is free.