SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2002.11537 · NeurIPS · 2020

ICE-BeeM: Identifiable Conditional Energy-Based Deep Models Based on Nonlinear ICA

Ilyes Khemakhem, Ricardo Monti, Diederik Kingma, Aapo Hyvärinen

arXiv · PDF · Open in the Atlas

Code that ran

3 of 4 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 2 have no contract verdict; 1 failure of our instrument; 1 no recorded run), measured 2026-11-21; 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/2002.11537.json · badge: /api/ran/2002.11537.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
ilkhem/icebeem canonical (classifier label)
classifier basis: classifier_2025
2 ran · 1 no recorded run
copy not recorded community
not in the paper's declared repositories
1 ran · 0 no recorded run
FunctionStatusWhere it lives
CleanMLP Ran
ran on a synthesized input
ilkhem/icebeem/models/ebm.py
code served (permissive licence) · get_code("70262264820e836e")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned Tensor, ndim 2, dtype family float, shape (2, 3) 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 torch
from torch import nn

class CleanMLP(nn.Module):

    def __init__(self, input_size, hidden_size, n_hidden, output_size, activation='lrelu', batch_norm=False):
        super().__init__()
        self.input_size = input_size
        self.output_size = output_size
        self.hidden_size = hidden_size
        self.n_hidden = n_hidden
        self.activation = activation
        self.batch_norm = batch_norm
        if activation == 'lrelu':
            act = nn.LeakyReLU(0.2, inplace=True)
        elif activation == 'relu':
            act = nn.ReLU()
        else:
            raise ValueError('wrong activation')
        if n_hidden == 0:
            modules = [nn.Linear(input_size, output_size)]
        else:
            modules = [nn.Linear(input_size, hidden_size), act] + batch_norm * [nn.BatchNorm1d(hidden_size)]
        for i in range(n_hidden - 1):
            modules += [nn.Linear(hidden_size, hidden_size), act] + batch_norm * [nn.BatchNorm1d(hidden_size)]
        modules += [nn.Linear(hidden_size, output_size)]
        self.net = nn.Sequential(*modules)

    def forward(self, x, y=None):
        return self.net(x)


INIT_ARGS = {
    'input_size': 4,
    'hidden_size': 8,
    'n_hidden': 2,
    'output_size': 3,
    'activation': 'lrelu',
    'batch_norm': False,
}

ARGS = {
    'x': torch.randn(2, 4),
}
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (2, 3) — shape (2, 3) 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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
UnnormalizedConditialEBM Ran
ran on a synthesized input
ilkhem/icebeem/models/ebm.py
code served (permissive licence) · get_code("08d0f658c6196bcf")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned Tensor, ndim 1, dtype family float, 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 torch
import torch.nn as nn
import torch.nn.functional as F

class CleanMLP(nn.Module):

    def __init__(self, input_size, hidden_size, n_hidden, output_size, activation='lrelu', batch_norm=False):
        super().__init__()
        self.input_size = input_size
        self.output_size = output_size
        self.hidden_size = hidden_size
        self.n_hidden = n_hidden
        self.activation = activation
        self.batch_norm = batch_norm
        if activation == 'lrelu':
            act = nn.LeakyReLU(0.2, inplace=True)
        elif activation == 'relu':
            act = nn.ReLU()
        else:
            raise ValueError('wrong activation')
        if n_hidden == 0:
            modules = [nn.Linear(input_size, output_size)]
        else:
            modules = [nn.Linear(input_size, hidden_size), act] + batch_norm * [nn.BatchNorm1d(hidden_size)]
        for i in range(n_hidden - 1):
            modules += [nn.Linear(hidden_size, hidden_size), act] + batch_norm * [nn.BatchNorm1d(hidden_size)]
        modules += [nn.Linear(hidden_size, output_size)]
        self.net = nn.Sequential(*modules)

    def forward(self, x, y=None):
        return self.net(x)

class UnnormalizedConditialEBM(nn.Module):

    def __init__(self, input_size, hidden_size, n_hidden, output_size, condition_size, activation='lrelu', augment=False, positive=False):
        super().__init__()
        self.input_size = input_size
        self.output_size = output_size
        self.hidden_size = hidden_size
        self.cond_size = condition_size
        self.n_hidden = n_hidden
        self.activation = activation
        self.augment = augment
        self.positive = positive
        self.f = CleanMLP(input_size, hidden_size, n_hidden, output_size, activation=activation)
        self.g = nn.Linear(condition_size, output_size, bias=False)

    def forward(self, x, y):
        fx, gy = (self.f(x).view(-1, self.output_size), self.g(y))
        if self.positive:
            fx = F.relu(fx)
            gy = F.relu(gy)
        if self.augment:
            return torch.einsum('bi,bi->b', [fx, gy]) + torch.einsum('bi,bi->b', [fx.pow(2), gy.pow(2)])
        else:
            return torch.einsum('bi,bi->b', [fx, gy])

INIT_ARGS = {
    'input_size': 4,
    'hidden_size': 8,
    'n_hidden': 2,
    'output_size': 3,
    'condition_size': 5,
    'activation': 'lrelu',
    'augment': False,
    'positive': False,
}

ARGS = {
    'x': torch.randn(2, 4),
    'y': torch.randn(2, 5),
}

if __name__ == '__main__':
    model = UnnormalizedConditialEBM(**INIT_ARGS)
    output = model(**ARGS)
    print(f"Output shape: {output.shape}")
    print(f"Output: {output}")
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 1, dtype family float, 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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
dict2namespace Ran
ran; our contract draft was wrong, not the code
this paper's copy was not recorded; identical code first harvested from leaveslei/obd
pointer only; the stored copy get_code serves, from leaveslei/obd, is pointer only (licence: NONE) · get_code("bd1f17e427bf51a5")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

ConditionalEBM Not yet run
unverified (harvested, no recorded run)
ilkhem/icebeem/models/ebm.py
code served (permissive licence) · get_code("0b97b2463f427a8f")
How it ran

no invocation recorded.

Also present in other papers' code

Of 4 samples lifted from this paper's linked repositories, 4 carry a function identity and 0 do not. Of the 4 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, 1 later) and 3 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
dict2namespace 28 other papers
2 dated earlier, 26 later
1907.04809 (2019-07)
Variational Autoencoders and Nonlinear ICA: A Unifying Framework
1912.00589 (2019-12)
Flow Contrastive Estimation of Energy-Based Models
2202.13711 (2022-02)
Evaluating the Adversarial Robustness of Adaptive Test-time Defenses
and 25 more not named here

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 2026-11-21.

Repositories linked to this paper

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

Abstract

We consider the identifiability theory of probabilistic models and establish sufficient conditions under which the representations learned by a very broad family of conditional energy-based models are unique in function space, up to a simple transformation. In our model family, the energy function is the dot-product between two feature extractors, one for the dependent variable, and one for the conditioning variable. We show that under mild conditions, the features are unique up to scaling and permutation. Our results extend recent developments in nonlinear ICA, and in fact, they lead to an important generalization of ICA models. In particular, we show that our model can be used for the estimation of the components in the framework of Independently Modulated Component Analysis (IMCA), a new generalization of nonlinear ICA that relaxes the independence assumption. A thorough empirical study shows that representations learned by our model from real-world image datasets are identifiable, and improve performance in transfer learning and semi-supervised learning tasks.

For agents

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

get_harvested_code_for_paper("2002.11537")
get_code_for_paper("2002.11537")
have("2002.11537")

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

curl https://syntology.ai/api/ran/2002.11537.json

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

[![Syntology run record](https://syntology.ai/api/ran/2002.11537.svg)](https://syntology.ai/paper/2002.11537)

Connect an agent — have() is free.