SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1909.03601 · 2019

Unbiased Recommender Learning from Missing-Not-At-Random Implicit Feedback

arXiv · PDF · Open in the Atlas

Code that ran

9 of 13 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; 1 honoured the contract we drafted, 0 violated it and 8 have no contract verdict; 0 failures of our instrument; 4 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/1909.03601.json · badge: /api/ran/1909.03601.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
usaito/unbiased-implicit-rec-real canonical (classifier label)
classifier basis: census
5 ran · 2 no recorded run
usaito/unbiased-implicit-rec canonical (classifier label)
classifier basis: census
4 ran · 2 no recorded run
FunctionStatusWhere it lives
average_precision_at_k Ran
ran on a synthesized input
usaito/unbiased-implicit-rec/src/utils/metrics.py
code served (permissive licence) · get_code("e91e3631a47d3246")
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 float64, ndim 0, dtype family float; 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

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

# Create sample data for average_precision_at_k
# y_true: binary labels indicating whether items are relevant
y_true = np.array([1, 0, 1, 0, 1, 0])

# y_score: relevance scores for ranking
y_score = np.array([0.9, 0.2, 0.8, 0.1, 0.7, 0.3])

# k: number of top items to consider
k = 3

ARGS = {
    "y_true": y_true,
    "y_score": y_score,
    "k": k
}
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 float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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
average_precision_at_k Ran
ran on a synthesized input
usaito/unbiased-implicit-rec-real/src/evaluate/metrics.py
code served (permissive licence) · get_code("d45b396101abb598")
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 float64, ndim 0, dtype family float; 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

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

# Create small test arrays
# y_true: binary labels (1 = relevant, 0 = not relevant)
y_true = np.array([1, 0, 1, 0, 1])

# y_score: relevance scores (higher = more relevant)
y_score = np.array([0.9, 0.2, 0.8, 0.1, 0.7])

# k: cutoff for top-k evaluation
k = 3

# pscore: optional propensity scores (probability of item being observed)
pscore = np.array([0.5, 0.3, 0.8, 0.2, 0.6])

ARGS = {
    "y_true": y_true,
    "y_score": y_score,
    "k": k,
    "pscore": pscore,
}
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 float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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
dcg_at_k Ran
ran on a synthesized input
usaito/unbiased-implicit-rec/src/utils/metrics.py
code served (permissive licence) · get_code("d7a1c19d49666f27")
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 float64, ndim 0, dtype family float; 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

# Construct arguments for dcg_at_k function
# The function expects:
# - y_true: np.ndarray of binary relevance labels
# - y_score: np.ndarray of predicted scores
# - k: int, the cutoff position

# Create sample data
y_true = np.array([1, 0, 1, 1, 0, 1])
y_score = np.array([0.9, 0.5, 0.8, 0.7, 0.3, 0.95])
k = 3

ARGS = {
    "y_true": y_true,
    "y_score": y_score,
    "k": k
}
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 float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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
dcg_at_k Ran
ran on a synthesized input
usaito/unbiased-implicit-rec-real/src/evaluate/metrics.py
code served (permissive licence) · get_code("5c67f0bf32203ff5")
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 float64, ndim 0, dtype family float; 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

# Create sample data for the function
y_true = np.array([1, 0, 1, 0, 1])
y_score = np.array([0.9, 0.2, 0.8, 0.3, 0.7])
k = 3
pscore = np.array([0.8, 0.5, 0.9, 0.4, 0.7])

ARGS = {
    "y_true": y_true,
    "y_score": y_score,
    "k": k,
    "pscore": pscore,
}
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 float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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
get_row Ran
ran on a synthesized input
usaito/unbiased-implicit-rec-real/src/models/expomf.py
code served (permissive licence) · get_code("6119496ab503be16")
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, ndim 2, dtype family float; 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 scipy.sparse
import numpy as np

# Create a small sparse CSR matrix
data = np.array([1.0, 2.0, 3.0, 4.0, 5.0])
indices = np.array([0, 2, 1, 3, 2])
indptr = np.array([0, 2, 3, 5])  # 3 rows

Y = scipy.sparse.csr_matrix((data, indices, indptr), shape=(3, 4))

ARGS = {
    "Y": Y,
    "i": 1
}
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, ndim 2, dtype family float — a tuple 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
precision_at_k Ran
ran on a synthesized input
usaito/unbiased-implicit-rec/src/utils/metrics.py
code served (permissive licence) · get_code("67693b649c55a340")
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 float64, ndim 0, dtype family float; 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

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

# Create sample data for precision_at_k
# y_true: binary labels (0 or 1) indicating relevance
y_true = np.array([1, 0, 1, 0, 1, 0])

# y_score: predicted scores/probabilities
y_score = np.array([0.9, 0.8, 0.7, 0.6, 0.5, 0.4])

# k: number of top items to consider
k = 3

ARGS = {
    "y_true": y_true,
    "y_score": y_score,
    "k": k,
}
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 float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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
recall_at_k Ran
ran on a synthesized input
usaito/unbiased-implicit-rec-real/src/evaluate/metrics.py
code served (permissive licence) · get_code("e58bf635478c876d")
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 float64, ndim 0, dtype family float; 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

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

# Create small test arrays for recall_at_k function
# y_true: binary labels indicating relevant items (shape: 10,)
y_true = np.array([1, 0, 1, 0, 1, 0, 0, 1, 0, 1])

# y_score: prediction scores for each item (shape: 10,)
y_score = np.array([0.9, 0.2, 0.85, 0.1, 0.95, 0.3, 0.15, 0.88, 0.05, 0.92])

# k: number of top items to consider
k = 5

# pscore: propensity scores (optional, can be None or an array)
# If provided, should have same shape as y_true and y_score
pscore = np.array([0.5, 0.6, 0.7, 0.8, 0.5, 0.6, 0.7, 0.8, 0.5, 0.6])

ARGS = {
    "y_true": y_true,
    "y_score": y_score,
    "k": k,
    "pscore": pscore
}
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 float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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
recompute_factors Ran
ran on a synthesized input
usaito/unbiased-implicit-rec-real/src/models/expomf.py
code served (permissive licence) · get_code("1ec2f637ffe2951a")
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 (6, 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 scipy.sparse as sp
from math import sqrt

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

# Create small sparse matrix Y (m x n)
# m=6 (users), n=5 (items)
m, n = 6, 5
row = np.array([0, 1, 2, 3, 4, 5, 0, 1, 2])
col = np.array([0, 1, 2, 3, 4, 0, 1, 2, 3])
data = np.array([1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0])
Y = sp.csr_matrix((data, (row, col)), shape=(m, n), dtype=np.float64)

# X: factor matrix for items (n x f), where f=4 is latent dimension
f = 4
X = np.random.randn(n, f).astype(np.float64)

# X_old: factor matrix for users (m x f)
X_old = np.random.randn(m, f).astype(np.float64)

# Regularization parameters
lam = 0.1  # ridge regularization
lam_y = 1.0  # exposure regularization

# mu: exposure parameter, can be scalar or per-user
mu = np.full(m, 0.5, dtype=np.float64)  # per-user exposure

# Parallelization
n_jobs = 1

# Batch size
batch_size = 2

ARGS = {
    "X": X,
    "X_old": X_old,
    "Y": Y,
    "lam": lam,
    "lam_y": lam_y,
    "mu": mu,
    "n_jobs": n_jobs,
    "batch_size": batch_size
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
joblib, 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 2, dtype family float, shape (6, 4) — shape (6, 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
sigmoid Ran
ran, honoured the contract we drafted
usaito/unbiased-implicit-rec/src/utils/data_generator.py
code served (permissive licence) · get_code("ebaf172fd8f34860")
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

# Create a small numpy array to pass to sigmoid
x = np.array([[1.0, 2.0, -1.0, 0.5],
              [-2.0, 0.0, 3.0, -0.5]])

ARGS = {
    "x": x
}
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 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
a_row_batch Not yet run
unverified (harvested, no recorded run)
usaito/unbiased-implicit-rec-real/src/models/expomf.py
code served (permissive licence) · get_code("7975cad450f46fde")
How it ran

no invocation recorded.

generate_sys_data Not yet run
unverified (harvested, no recorded run)
usaito/unbiased-implicit-rec/src/utils/data_generator.py
code served (permissive licence) · get_code("9be5e439e2a3c849")
How it ran

no invocation recorded.

load_movielens Not yet run
unverified (harvested, no recorded run)
usaito/unbiased-implicit-rec/src/utils/data_generator.py
code served (permissive licence) · get_code("bd43d52fb94a700e")
How it ran

no invocation recorded.

tocsr Not yet run
unverified (harvested, no recorded run)
usaito/unbiased-implicit-rec-real/src/trainer.py
code served (permissive licence) · get_code("193158bf5121b5b3")
How it ran

no invocation recorded.

Also present in other papers' code

Of 13 samples lifted from this paper's linked repositories, 13 carry a function identity and 0 do not. Of the 13 distinct functions with an identity, 1 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 1 later) and 12 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
sigmoid 1 other paper
1 later
aaai_29453 (2024 (year only))
Comprehensive View Embedding Learning for Single-Cell Multimodal Integration

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

Recommender systems widely use implicit feedback such as click data because of its general availability. Although the presence of clicks signals the users' preference to some extent, the lack of such clicks does not necessarily indicate a negative response from the users, as it is possible that the users were not exposed to the items (positive-unlabeled problem). This leads to a difficulty in predicting the users' preferences from implicit feedback. Previous studies addressed the positive-unlabeled problem by uniformly upweighting the loss for the positive feedback data or estimating the confidence of each data having relevance information via the EM-algorithm. However, these methods failed to address the missing-not-at-random problem in which popular or frequently recommended items are more likely to be clicked than other items even if a user does not have a considerable interest in them. To overcome these limitations, we first define an ideal loss function to be optimized to realize recommendations that maximize the relevance and propose an unbiased estimator for the ideal loss. Subsequently, we analyze the variance of the proposed unbiased estimator and further propose a clipped estimator that includes the unbiased estimator as a special case. We demonstrate that the clipped estimator is expected to improve the performance of the recommender system, by considering the bias-variance trade-off. We conduct semi-synthetic and real-world experiments and demonstrate that the proposed method largely outperforms the baselines. In particular, the proposed method works better for rare items that are less frequently observed in the training data. The findings indicate that the proposed method can better achieve the objective of recommending items with the highest relevance.

For agents

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

get_harvested_code_for_paper("1909.03601")
get_code_for_paper("1909.03601")
have("1909.03601")

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

curl https://syntology.ai/api/ran/1909.03601.json

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

[![Syntology run record](https://syntology.ai/api/ran/1909.03601.svg)](https://syntology.ai/paper/1909.03601)

Connect an agent — have() is free.