SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1808.03233 · 2018

OBOE: Collaborative Filtering for AutoML Model Selection

arXiv · PDF · Open in the Atlas

Code that ran

9 of 11 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 9 have no contract verdict; 0 failures of our instrument; 2 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; 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/1808.03233.json · badge: /api/ran/1808.03233.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
udellgroup/oboe canonical (classifier label)
classifier basis: census
9 ran · 2 no recorded run
FunctionStatusWhere it lives
approx_matrix_rank Ran
ran on a synthesized input
udellgroup/oboe/oboe/linalg.py
code served (permissive licence) · get_code("94ee8835ad26e950")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned int, ndim 0, dtype family int, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape 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 numpy as np

# Create a test matrix with known rank structure
# Use a 5x4 matrix to have distinct dimensions
np.random.seed(0)
a = np.random.randn(5, 4)

# Set up the arguments dictionary
ARGS = {
    "a": a,
    "threshold": 0.03
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type int, ndim 0, dtype family int, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
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
extract_columns Ran
ran on a synthesized input
udellgroup/oboe/oboe/util.py
code served (permissive licence) · get_code("c85fdd80f7694c19")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned DataFrame. 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 pandas as pd
import numpy as np

# Create a sample error matrix DataFrame
# The columns need to be strings that can be eval'd into dicts with 'algorithm' keys
data = {
    "{'algorithm': 'RandomForest', 'n_estimators': 100, 'max_depth': 5}": [0.1, 0.2, 0.15],
    "{'algorithm': 'RandomForest', 'n_estimators': 50, 'max_depth': 3}": [0.12, 0.22, 0.17],
    "{'algorithm': 'SVM', 'kernel': 'rbf', 'C': 1.0}": [0.2, 0.25, 0.22],
    "{'algorithm': 'SVM', 'kernel': 'linear', 'C': 0.5}": [0.18, 0.23, 0.20],
    "{'algorithm': 'LogisticRegression', 'penalty': 'l2', 'C': 1.0}": [0.15, 0.18, 0.16],
}

df = pd.DataFrame(data)

# Define algorithms to extract
algorithms = ['RandomForest', 'SVM']

# Define hyperparameters constraints
hyperparameters = {
    'RandomForest': {
        'n_estimators': [100, 200],
        'max_depth': [5, 10]
    },
    'SVM': {
        'kernel': ['rbf', 'poly'],
        'C': [1.0, 2.0]
    }
}

ARGS = {
    'df': df,
    'algorithms': algorithms,
    'hyperparameters': hyperparameters
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DataFrame — a DataFrame output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataFrame output the recorded ndim 2 and dtype family float describe that coercion, not the output, and are not 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
get_omega Ran
ran on a synthesized input
udellgroup/oboe/oboe/util.py
code served (permissive licence) · get_code("7746c80e86bb278d")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 2, dtype family float, shape (2, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape 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 numpy as np
import torch

torch.manual_seed(0)

# Create a small numpy array with some NaN values to test the function
tensor = np.array([
    [1.0, 2.0, np.nan, 4.0],
    [5.0, np.nan, 7.0, 8.0]
])

ARGS = {
    "tensor": tensor
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (2, 4) — shape (2, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
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
greedy_stepwise_selection_with_time Ran
ran on a synthesized input
udellgroup/oboe/oboe/experiment_design.py
code served (permissive licence) · get_code("b2dbc591d7263abb")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family int; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim 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 numpy as np

np.random.seed(0)

# Y should be a 2D array where shape is (n_samples, n_cols)
# The function accesses Y[:, initialization] and Y[:, i]
n_samples = 5
n_cols = 8
Y = np.random.randn(n_samples, n_cols)

# t should be a 1D array with length equal to n_cols
# It represents time costs for each column
t = np.array([0.5, 0.3, 0.4, 0.6, 0.2, 0.5, 0.3, 0.4])

# initialization should be a list/array of indices to start with
initialization = [0, 1]

# t_elapsed is the time already elapsed
t_elapsed = 0.0

# t_max is the maximum time allowed
t_max = 3.0

# idx_to_exclude is optional list of indices to exclude from selection
idx_to_exclude = [7]

# verbose flag for printing
verbose = False

ARGS = {
    'Y': Y,
    't': t,
    'initialization': initialization,
    't_elapsed': t_elapsed,
    't_max': t_max,
    'idx_to_exclude': idx_to_exclude,
    'verbose': verbose
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family int — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
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
pivot_columns Ran
ran on a synthesized input
udellgroup/oboe/oboe/linalg.py
code served (permissive licence) · get_code("8f03823b0f3724b0")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 1, dtype family int, shape (2,) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape 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 numpy as np

# Create a small test matrix for the function
# Using a 4x3 matrix with distinct dimensions
np.random.seed(42)
a_matrix = np.random.randn(4, 3)

# Define arguments for pivot_columns
# The function requires exactly one of 'threshold' or 'rank' to be specified
ARGS = {
    'a': a_matrix,
    'rank': 2,
    'threshold': None
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 1, dtype family int, shape (2,) — shape (2,) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
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
pivot_columns_time Ran
ran on a synthesized input
udellgroup/oboe/oboe/experiment_design.py
code served (permissive licence) · get_code("ea9736013b7c72b0")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) 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 numpy as np
from scipy.linalg import qr

# Set random seed for reproducibility
np.random.seed(0)

# Create a small matrix A with shape (5, 4)
A = np.random.randn(5, 4)

# Create time array t with length 4 (matching number of columns)
t = np.array([0.1, 0.2, 0.15, 0.25])

# Set t_max large enough to allow some columns to be selected
t_max = 1.0

# Set rank to 2
rank = 2

# Create columns_to_avoid list (empty or with some indices)
columns_to_avoid = [0]

# threshold is None since rank is specified
threshold = None

# column_picking_threshold should be 'rank' or 'size'
column_picking_threshold = 'rank'

ARGS = {
    'A': A,
    't': t,
    't_max': t_max,
    'rank': rank,
    'columns_to_avoid': columns_to_avoid,
    'threshold': threshold,
    'column_picking_threshold': column_picking_threshold
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — 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
pre_process Ran
ran on a synthesized input
udellgroup/oboe/oboe/preprocessing.py
code served (permissive licence) · get_code("fdc04999e73c7039")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) 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 numpy as np
from sklearn.preprocessing import scale
from sklearn.compose import ColumnTransformer
from sklearn.preprocessing import OneHotEncoder
from sklearn.impute import SimpleImputer

# Create sample raw data: 5 samples, 4 features
# Mix of categorical and numeric features
np.random.seed(0)
raw_data = np.array([
    [1.0, 'A', 10.5, 'X'],
    [2.0, 'B', np.nan, 'Y'],
    [np.nan, 'A', 20.3, 'X'],
    [3.0, 'C', 15.1, np.nan],
    [1.0, 'B', 18.9, 'Y']
], dtype=object)

# Convert numeric columns to float, keep categorical as strings
raw_data_clean = np.empty((5, 4), dtype=object)
raw_data_clean[:, 0] = raw_data[:, 0].astype(float)
raw_data_clean[:, 1] = raw_data[:, 1]
raw_data_clean[:, 2] = raw_data[:, 2].astype(float)
raw_data_clean[:, 3] = raw_data[:, 3]

# Boolean list indicating which features are categorical
# Feature 0: numeric, Feature 1: categorical, Feature 2: numeric, Feature 3: categorical
categorical = [False, True, False, True]

ARGS = {
    'raw_data': raw_data_clean,
    'categorical': categorical,
    'impute': True,
    'standardize': True,
    'one_hot_encode': True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, sklearn — 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
solve Ran
ran on a synthesized input
udellgroup/oboe/oboe/experiment_design.py
code served (permissive licence) · get_code("35a4536bc536046d")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned ndarray, ndim 1, dtype family float, shape (5,) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape 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 numpy as np

# Set random seed for reproducibility
np.random.seed(0)

# Create small test data
# t_predicted: 1-d array of predicted runtimes for each model setting
n_models = 5
t_predicted = np.array([0.5, 1.2, 0.8, 1.5, 0.3])

# t_max: maximum runtime budget
t_max = 2.0

# n_cores: number of cores available
n_cores = 4

# Y: matrix representing latent variable weights of error matrix
# Shape: (n_latent, n_models) where n_latent < n_models for a meaningful problem
n_latent = 3
Y = np.random.randn(n_latent, n_models)

# scalarization: method for experimental design ('D', 'A', or 'E')
scalarization = 'D'

ARGS = {
    't_predicted': t_predicted,
    't_max': t_max,
    'n_cores': n_cores,
    'Y': Y,
    'scalarization': scalarization,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 1, dtype family float, shape (5,) — shape (5,) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
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
tucker_on_error_tensor Ran
ran on a synthesized input
udellgroup/oboe/oboe/util.py
code served (permissive licence) · get_code("f2f0dda34422aeac")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) 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 numpy as np

# Create a small error tensor with some NaN values
# Shape: (3, 4, 5, 2, 3, 2) - 6D tensor matching the default rank length
np.random.seed(0)
error_tensor = np.random.randn(3, 4, 5, 2, 3, 2)
# Add some NaN values to make it realistic
error_tensor[0, 1, 2, 0, 1, 0] = np.nan
error_tensor[1, 2, 3, 1, 2, 1] = np.nan
error_tensor[2, 0, 1, 0, 0, 0] = np.nan

ARGS = {
    'error_tensor': error_tensor,
    'rank': [2, 2, 2, 2, 2, 2],  # Smaller rank for faster computation on small tensor
    'save_results': False,
    'save_path': '',
    'verbose': False
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, sklearn, tensorly — 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
approx_tensor_rank Not yet run
unverified (harvested, no recorded run)
udellgroup/oboe/oboe/linalg.py
code served (permissive licence) · get_code("7c1688175e9658ee")
How it ran

no invocation recorded.

parse_args Not yet run
unverified (harvested, no recorded run)
udellgroup/oboe/oboe/generate_vector.py
code served (permissive licence) · get_code("3f6277b4c8c18f9f")
How it ran

no invocation recorded.

Also present in other papers' code

Of 11 samples lifted from this paper's linked repositories, 11 carry a function identity and 0 do not. Of the 11 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 11 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.

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

Algorithm selection and hyperparameter tuning remain two of the most challenging tasks in machine learning. Automated machine learning (AutoML) seeks to automate these tasks to enable widespread use of machine learning by non-experts. This paper introduces OBOE, a collaborative filtering method for time-constrained model selection and hyperparameter tuning. OBOE forms a matrix of the cross-validated errors of a large number of supervised learning models (algorithms together with hyperparameters) on a large number of datasets, and fits a low rank model to learn the low-dimensional feature vectors for the models and datasets that best predict the cross-validated errors. To find promising models for a new dataset, OBOE runs a set of fast but informative algorithms on the new dataset and uses their cross-validated errors to infer the feature vector for the new dataset. OBOE can find good models under constraints on the number of models fit or the total time budget. To this end, this paper develops a new heuristic for active learning in time-constrained matrix completion based on optimal experiment design. Our experiments demonstrate that OBOE delivers state-of-the-art performance faster than competing approaches on a test bed of supervised learning problems. Moreover, the success of the bilinear model used by OBOE suggests that AutoML may be simpler than was previously understood.

For agents

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

get_harvested_code_for_paper("1808.03233")
get_code_for_paper("1808.03233")
have("1808.03233")

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

curl https://syntology.ai/api/ran/1808.03233.json

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

[![Syntology run record](https://syntology.ai/api/ran/1808.03233.svg)](https://syntology.ai/paper/1808.03233)

Connect an agent — have() is free.