SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2104.05353 · ICLR · 2021

Sparse Coding Frontend for Robust Neural Networks

Can Bakiskan, Metehan Cekic, Ahmet Sezer, Upamanyu Madhow

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 (2 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 2026-11-22; 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/2104.05353.json · badge: /api/ran/2104.05353.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
canbakiskan/sparse_coding_frontend canonical (classifier label)
classifier basis: classifier_2025
3 ran · 1 no recorded run
FunctionStatusWhere it lives
encoder Ran
ran on a synthesized input · constructed an object; no result was computed
canbakiskan/sparse_coding_frontend/src/models/encoder.py
code served (permissive licence) · get_code("4e4f228b3b9b546c")
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 an instance of its own class, encoder: the run constructed the object and did not call it, so no result was computed. 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 torch
from torch import nn

torch.manual_seed(0)

# Create a minimal config object that matches the args structure
class DictConfig:
    def __init__(self, **kwargs):
        for key, value in kwargs.items():
            setattr(self, key, value)

class DefenseConfig:
    def __init__(self):
        self.assumed_budget = 0.1
        self.train_noise_gamma = 0.5
        self.test_noise_gamma = 0.3
        self.frontend_arch = "basic"  # minimal arch without extra features
        self.patch_size = 3
        self.stride = 1
        self.dropout_p = 0.2
        self.top_T = 5
        self.activation_beta = 0.1

class DictionaryConfig:
    def __init__(self):
        self.nb_atoms = 4
        self.dim = 9  # patch_size^2 * channels = 3*3*1

class AblationConfig:
    def __init__(self):
        self.no_dictionary = True

class AdvTestingConfig:
    def __init__(self):
        self.top_T_backward = 'default'
        self.dropout_backward = 'identity'
        self.activation_backward_steepness = 0.0
        self.top_U = 3

class Args:
    def __init__(self):
        self.defense = DefenseConfig()
        self.dictionary = DictionaryConfig()
        self.ablation = AblationConfig()
        self.adv_testing = AdvTestingConfig()
        self.seed = 42

ARGS = {
    'args': Args()
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type encoder — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a encoder output the recorded ndim 0 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
extract_patches Ran
ran; our contract draft was wrong, not the code
canbakiskan/sparse_coding_frontend/src/learn_patch_dict.py
code served (permissive licence) · get_code("5b1bb8f496d81deb")
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) and returned ndarray, ndim 4, dtype family float, shape (12, 3, 3, 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
import numpy as np

torch.manual_seed(0)

# Test with numpy array - NHWC format
images_np = np.random.randn(2, 4, 5, 3).astype(np.float32)
patch_shape = (3, 3, 3)
stride = 1
in_order = 'NHWC'
out_order = 'NHWC'

ARGS = {
    'images': images_np,
    'patch_shape': patch_shape,
    'stride': stride,
    'in_order': in_order,
    'out_order': out_order
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, sklearn, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 4, dtype family float, shape (12, 3, 3, 3) — shape (12, 3, 3, 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: none by the harness; any seed is in the setup block
noisy Ran
ran on a synthesized input · constructed an object; no result was computed
canbakiskan/sparse_coding_frontend/src/models/encoder.py
code served (permissive licence) · get_code("ad11a2a34ae8d395")
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 an instance of its own class, noisy: the run constructed the object and did not call it, so no result was computed. 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 torch

torch.manual_seed(0)

class DefenseConfig:
    def __init__(self):
        self.assumed_budget = 0.5
        self.train_noise_gamma = 0.1
        self.test_noise_gamma = 0.05

class Args:
    def __init__(self):
        self.defense = DefenseConfig()

# Create a noisy instance to inspect attributes needed
args = Args()

# The noisy class needs:
# - self.train_noise_level (computed in __init__)
# - self.test_noise_level (computed in __init__)
# - self.training (attribute, used in add_noise)
# - self.l1_norms (attribute, used in add_noise as a tensor)

# Create a mock noisy object with required attributes
class NoisyMock:
    def __init__(self, args):
        self.train_noise_level = args.defense.assumed_budget * args.defense.train_noise_gamma
        self.test_noise_level = args.defense.assumed_budget * args.defense.test_noise_gamma
        self.training = True
        self.l1_norms = torch.tensor([0.1, 0.2, 0.15, 0.25])  # 4 channels
    
    def add_noise(self, x, seed=None):
        if seed:
            torch.manual_seed(seed)
        noise = torch.rand_like(x)
        noise = noise / torch.norm(noise, p=1, dim=(2, 3)).unsqueeze(-1).unsqueeze(-1)
        if self.training:
            noise = noise * self.train_noise_level * self.l1_norms.view(1, -1, 1, 1)
        else:
            noise = noise * self.test_noise_level * self.l1_norms.view(1, -1, 1, 1)
        return x + noise

ARGS = {
    "args": Args()
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type noisy — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a noisy output the recorded ndim 0 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
dropout Not yet run
unverified (harvested, no recorded run)
canbakiskan/sparse_coding_frontend/src/models/encoder.py
code served (permissive licence) · get_code("25c2d65fbcdc6bc7")
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, 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 4 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

Deep Neural Networks are known to be vulnerable to small, adversarially crafted, perturbations. The current most effective defense methods against these adversarial attacks are variants of adversarial training. In this paper, we introduce a radically different defense trained only on clean images: a sparse coding based front end which significantly attenuates adversarial attacks before they reach the classifier. We evaluate our defense on CIFAR-10 dataset under a wide range of attack types (including ∞ , 2 , and 1 bounded attacks), demonstrating its promise as a general-purpose approach for defense.

For agents

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

get_harvested_code_for_paper("2104.05353")
get_code_for_paper("2104.05353")
have("2104.05353")

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

curl https://syntology.ai/api/ran/2104.05353.json

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

[![Syntology run record](https://syntology.ai/api/ran/2104.05353.svg)](https://syntology.ai/paper/2104.05353)

Connect an agent — have() is free.