SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2110.04176 · 2021

PHNNs: Lightweight Neural Networks via Parameterized Hypercomplex Convolutions

arXiv · PDF · Open in the Atlas

Code that ran

28 of 34 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 23 have no contract verdict; 4 failures of our instrument; 6 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/2110.04176.json · badge: /api/ran/2110.04176.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
elegan23/hypernets canonical (classifier label)
classifier basis: classifier_2025
0 ran · 2 no recorded run
ispamm/hi2i community (classifier label)
classifier basis not recorded on the link
19 ran · 2 no recorded run
eleGAN23/QVAE community (classifier label)
classifier basis not recorded on the link
7 ran · 0 no recorded run
eleGAN23/QGAN community (classifier label)
classifier basis not recorded on the link
2 ran · 2 no recorded run
FunctionStatusWhere it lives
QImproper_signals1 Ran
ran on a synthesized input
eleGAN23/QVAE/entropy/datasets.py
code served (permissive licence) · get_code("b9ef9bade0c9f9c8")
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 1, 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 torch

torch.manual_seed(0)

ARGS = {
    "len_train": 2,
    "len_val": 2,
    "len_test": 2,
    "n": 4,
}
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 tuple, ndim 1, 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
QProper_signals1 Ran
ran on a synthesized input
eleGAN23/QVAE/entropy/datasets.py
code served (permissive licence) · get_code("6760e9b0a1a4eea4")
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 1, 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 torch

torch.manual_seed(0)

ARGS = {
    "len_train": 2,
    "len_val": 2,
    "len_test": 2,
    "n": 4,
}
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 tuple, ndim 1, 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
QProper_signals2 Ran
ran on a synthesized input
eleGAN23/QVAE/entropy/datasets.py
code served (permissive licence) · get_code("d9d732f238cd29b2")
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 1, 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 torch
import numpy as np
from scipy.signal import lfilter

torch.manual_seed(0)
np.random.seed(0)

ARGS = {
    "len_train": 8,
    "len_val": 4,
    "len_test": 2,
    "n": 3
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, 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
adv_loss Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
ispamm/hi2i/core/solver.py
pointer only (licence: MIT); the stored copy get_code serves, from cjf8899/ReMix_Pytorch, is pointer only (licence: NONE) · get_code("16d563ee991cd39f")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from cjf8899/ReMix_Pytorch, and GitHub detected no licence in that repository when it was checked. 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.

compute_d_loss Ran
ran on a synthesized input
ispamm/hi2i/core/solver.py
code served (permissive licence) · get_code("55dc256ba1c04149")
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 torch
import torch.nn as nn

torch.manual_seed(0)

class SimpleDiscriminator(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 1)
    
    def forward(self, x, y):
        return self.linear(x)

class SimpleMappingNetwork(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(5, 4)
    
    def forward(self, z, y):
        return self.linear(z)

class SimpleStyleEncoder(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 4)
    
    def forward(self, x, y):
        return self.linear(x)

class SimpleGenerator(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(8, 4)
    
    def forward(self, x, s, masks=None):
        return self.linear(torch.cat([x, s], dim=1))

class SimpleNets:
    def __init__(self):
        self.discriminator = SimpleDiscriminator()
        self.mapping_network = SimpleMappingNetwork()
        self.style_encoder = SimpleStyleEncoder()
        self.generator = SimpleGenerator()

class SimpleArgs:
    def __init__(self):
        self.lambda_reg = 0.1

nets = SimpleNets()
args = SimpleArgs()
x_real = torch.randn(2, 4)
y_org = torch.randint(0, 2, (2,))
y_trg = torch.randint(0, 2, (2,))
z_trg = torch.randn(2, 5)
x_ref = None
masks = None

ARGS = {
    'nets': nets,
    'args': args,
    'x_real': x_real,
    'y_org': y_org,
    'y_trg': y_trg,
    'z_trg': z_trg,
    'x_ref': x_ref,
    'masks': masks
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
munch, torch — 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
compute_g_loss Ran
ran on a synthesized input
ispamm/hi2i/core/solver.py
code served (permissive licence) · get_code("d0c7af72cd606f53")
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 1, 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 torch
import torch.nn as nn
from munch import Munch

torch.manual_seed(0)

# Define minimal nn.Module implementations
class MappingNetwork(nn.Module):
    def forward(self, z, y):
        # Takes latent z and label y, returns style vector
        return torch.randn(z.size(0), 4)  # batch_size x style_dim

class StyleEncoder(nn.Module):
    def forward(self, x, y):
        # Takes image x and label y, returns style vector
        return torch.randn(x.size(0), 4)  # batch_size x style_dim

class Generator(nn.Module):
    def forward(self, x_real, s_trg, masks=None):
        # Takes real image and style, returns fake image
        return torch.randn_like(x_real)

class Discriminator(nn.Module):
    def forward(self, x_fake, y_trg):
        # Takes fake image and label, returns logits
        return torch.randn(x_fake.size(0), 1)

class FAN(nn.Module):
    def get_heatmap(self, x):
        # Takes image, returns mask
        return torch.randn(x.size(0), 1, x.size(2), x.size(3))

class Nets:
    def __init__(self):
        self.mapping_network = MappingNetwork()
        self.style_encoder = StyleEncoder()
        self.generator = Generator()
        self.discriminator = Discriminator()
        self.fan = FAN()

# Create minimal args object
class Args:
    def __init__(self):
        self.real = False
        self.w_hpf = 1
        self.lambda_sty = 1.0
        self.lambda_ds = 1.0
        self.lambda_cyc = 1.0

# Construct arguments
nets = Nets()
args = Args()
x_real = torch.randn(2, 3, 4, 4)  # batch=2, channels=3, spatial=4x4
y_org = torch.randint(0, 2, (2,))
y_trg = torch.randint(0, 2, (2,))
z_trg = torch.randn(2, 8)  # batch=2, latent_dim=8
z_trg2 = torch.randn(2, 8)
z_trgs = (z_trg, z_trg2)
x_refs = None
masks = None

ARGS = {
    'nets': nets,
    'args': args,
    'x_real': x_real,
    'y_org': y_org,
    'y_trg': y_trg,
    'z_trgs': z_trgs,
    'x_refs': x_refs,
    'masks': masks,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
munch, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, 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
convert_data_for_quaternion Ran
ran on a synthesized input
ispamm/hi2i/htorch/utils.py
code served (permissive licence) · get_code("470f567ca3e3fb39")
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 torch
import torchvision

torch.manual_seed(0)

# Create a small batch of RGB images
# Each image is 3 channels (RGB), height 4, width 4
batch = []
for i in range(2):
    # Create a random RGB image tensor (3, 4, 4)
    img = torch.randn(3, 4, 4)
    # Create a label (integer)
    label = i
    batch.append((img, label))

ARGS = {
    "batch": batch
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, torchvision — 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
convert_data_for_quaternion_STAR Ran
ran on a synthesized input
ispamm/hi2i/core/data_loader.py
code served (permissive licence) · get_code("cc1c58c9ca8943e1")
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 torch
import torchvision

torch.manual_seed(0)

# The function expects a batch where each element is a tuple of:
# (image1, image2, label)
# where image1 and image2 are RGB tensors (3 channels)

# Create small RGB images (3 channels, 4x4 spatial dimensions)
img1_1 = torch.randn(3, 4, 4)
img2_1 = torch.randn(3, 4, 4)
label1 = 0

img1_2 = torch.randn(3, 4, 4)
img2_2 = torch.randn(3, 4, 4)
label2 = 1

# Create batch as list of tuples
batch = [
    (img1_1, img2_1, label1),
    (img1_2, img2_2, label2)
]

ARGS = {
    "batch": batch
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, torchvision — 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
convert_data_for_quaternion_without_labels Ran
ran on a synthesized input
ispamm/hi2i/core/data_loader.py
code served (permissive licence) · get_code("fa6e591e0ad2d888")
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 Tensor, ndim 4, dtype family float, shape (2, 4, 4, 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 torch
import torchvision

torch.manual_seed(0)

# Create a batch of 2 RGB images with shape (3, 4, 5)
# 3 channels (RGB), 4x5 spatial dimensions
batch = [
    torch.randn(3, 4, 5),
    torch.randn(3, 4, 5),
]

ARGS = {
    "batch": batch,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, torchvision — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 4, dtype family float, shape (2, 4, 4, 5) — shape (2, 4, 4, 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
denormalize Ran
ran on a synthesized input
ispamm/hi2i/core/utils.py
code served (permissive licence) · get_code("4f63545a8e689c49")
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 Tensor, 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 torch

torch.manual_seed(0)

# The function denormalize takes a tensor x, applies (x + 1) / 2, and clamps to [0, 1]
# It expects x to be a tensor with a clamp_ method
x = torch.randn(2, 4)

ARGS = {
    "x": x
}
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 Tensor, 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
early_stopping Ran
ran on a synthesized input
eleGAN23/QVAE/train_vae.py
code served (permissive licence) · get_code("cb04204d244b0222")
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 1, 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 torch
import numpy as np

torch.manual_seed(0)

# The function early_stopping takes 5 parameters:
# - val_loss: a scalar loss value (float)
# - train_loss: a scalar loss value (float)
# - min_val_loss: a scalar representing minimum validation loss seen (float)
# - patience_epochs: an integer for patience threshold
# - epochs_no_improve: an integer counter for epochs without improvement

ARGS = {
    'val_loss': 0.5,
    'train_loss': 0.6,
    'min_val_loss': 0.4,
    'patience_epochs': 3,
    'epochs_no_improve': 1
}
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 tuple, ndim 1, 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
frechet_distance Ran
ran on a synthesized input
ispamm/hi2i/metrics/fid.py
code served (permissive licence) · get_code("bd115aa43372ad6b")
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
from scipy import linalg

# Create small covariance matrices and mean vectors for the Frechet distance calculation
np.random.seed(0)

# First distribution: mean vector and covariance matrix
mu = np.array([0.1, 0.2, 0.3, 0.4])
cov = np.array([
    [1.0, 0.1, 0.05, 0.02],
    [0.1, 0.9, 0.03, 0.01],
    [0.05, 0.03, 0.8, 0.04],
    [0.02, 0.01, 0.04, 0.7]
])

# Second distribution: mean vector and covariance matrix
mu2 = np.array([0.15, 0.25, 0.35, 0.45])
cov2 = np.array([
    [1.1, 0.12, 0.06, 0.03],
    [0.12, 0.95, 0.04, 0.02],
    [0.06, 0.04, 0.85, 0.05],
    [0.03, 0.02, 0.05, 0.75]
])

ARGS = {
    "mu": mu,
    "cov": cov,
    "mu2": mu2,
    "cov2": cov2
}
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 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_alphas Ran
ran on a synthesized input
ispamm/hi2i/core/utils.py
code served (permissive licence) · get_code("380e205f2b8ababc")
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 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

ARGS = {
    "start": -5,
    "end": 5,
    "step": 0.5,
    "len_tail": 10
}
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 float — 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
get_device Ran
ran on a synthesized input
ispamm/hi2i/htorch/quaternion.py
code served (permissive licence) · get_code("486935c1efc81482")
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 torch
import torch.nn as nn

torch.manual_seed(0)

class ScalarTensor:
    """Mock ScalarTensor class to demonstrate the get_device function"""
    def __init__(self, tensor):
        self._tensor = tensor
    
    def torch(self):
        """Return the underlying torch tensor"""
        return self._tensor

# Create a small tensor
small_tensor = torch.randn(2, 4)

# Create a ScalarTensor instance
scalar_tensor = ScalarTensor(small_tensor)

ARGS = {
    'self': scalar_tensor
}
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 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
get_i Ran
ran; our contract draft was wrong, not the code
eleGAN23/QVAE/utils/quaternion_ops.py
pointer only (licence: MIT); the stored copy get_code serves, from XinxiiChen/AGREE, is pointer only (licence: NONE) · get_code("01423bbbb5930bbd")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from XinxiiChen/AGREE, and GitHub detected no licence in that repository when it was checked. 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.

get_preds_fromhm Ran
ran on a synthesized input
ispamm/hi2i/core/wing.py
code served (permissive licence) · get_code("0dc6ba46b83190d8")
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 Tensor, ndim 3, dtype family float, shape (2, 4, 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

torch.manual_seed(0)

# Construct a heatmap tensor
# The function expects:
# - hm.size(0) = batch size
# - hm.size(1) = number of keypoints
# - hm.size(2) = height (must be > 1 for indexing at pY-1, pY, pY+1)
# - hm.size(3) = width (must be > 1 for indexing at pX-1, pX, pX+1)
# The function uses pX and pY with bounds checks for 0 < pX < 63 and 0 < pY < 63
# So we need size(2) and size(3) to be at least 64 to match the bounds check

hm = torch.randn(2, 4, 64, 64)

ARGS = {
    "hm": hm
}
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 Tensor, ndim 3, dtype family float, shape (2, 4, 2) — shape (2, 4, 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
get_r Ran
ran; our contract draft was wrong, not the code
eleGAN23/QVAE/utils/quaternion_ops.py
pointer only (licence: MIT); the stored copy get_code serves, from XinxiiChen/AGREE, is pointer only (licence: NONE) · get_code("250d2abbf60ffa77")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from XinxiiChen/AGREE, and GitHub detected no licence in that repository when it was checked. 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.

get_weight Ran
ran on a synthesized input
ispamm/hi2i/core/model.py
code served (permissive licence) · get_code("cf9f3072be35c325")
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 Tensor, ndim 2, dtype family float, shape (16, 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
from numpy.random import RandomState

torch.manual_seed(0)

# Analyze the function signature and usage:
# get_weight(n, in_f, out_f, kernel_size, criterion)
# - n: used in division (out_f // n), should be int
# - in_f: passed to quaternion_init as in_features, should be int
# - out_f: passed to quaternion_init as out_features, should be int
# - kernel_size: passed to quaternion_init, can be None, int, or tuple
# - criterion: passed to quaternion_init and used in if/else, string ('he' or 'glorot')

# The function calls quaternion_init and then uses torch operations
# It concatenates along dim=1 for 'he' or dim=0 for other criteria
# For proper shapes, we need to ensure the tensors can be concatenated

# Let's choose:
# - n: small int divisor of out_f
# - in_f: small int (e.g., 4)
# - out_f: small int divisible by n (e.g., 8)
# - kernel_size: None (simplest case, linear layer style)
# - criterion: 'glorot' (tests the squeeze and cat on dim=0 path)

ARGS = {
    'n': 2,
    'in_f': 4,
    'out_f': 8,
    'kernel_size': None,
    'criterion': 'glorot'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy, 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 (16, 4) — shape (16, 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
implements Ran
ran on a synthesized input
ispamm/hi2i/htorch/quaternion.py
code served (permissive licence) · get_code("346a1f46489131dd")
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 function. 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
import functools

torch.manual_seed(0)

# The function `implements` is a decorator factory that takes a torch_function
# as a parameter. It returns a decorator that registers functions in HANDLEdFUNCTIONS.
# 
# To call `implements`, we need to pass a torch function as the argument.
# We'll use a simple torch function like torch.add

ARGS = {
    "torch_function": torch.add
}
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 function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function 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
listdir Ran
ran on a synthesized input
ispamm/hi2i/core/data_loader.py
code served (permissive licence) · get_code("d07d6906ac8b165a")
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 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 tempfile
from pathlib import Path

# Create a temporary directory with some image files
tmpdir = tempfile.mkdtemp()
tmppath = Path(tmpdir)

# Create some dummy image files with the expected extensions
(tmppath / "image1.png").touch()
(tmppath / "image2.jpg").touch()
(tmppath / "image3.jpeg").touch()
(tmppath / "image4.JPG").touch()

# Also create a subdirectory with images to test rglob
subdir = tmppath / "subdir"
subdir.mkdir()
(subdir / "image5.png").touch()

ARGS = {
    "dname": tmpdir
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — 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 float — 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
moving_average_update Ran
ran, honoured the contract we drafted
eleGAN23/QGAN/utils/QBN_Vecchi2.py
pointer only (licence: MIT); the stored copy get_code serves, from asahi-research/QC-GAN, is pointer only (licence: NONE) · get_code("e2e18700dc6ed1a8")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from asahi-research/QC-GAN, and GitHub detected no licence in that repository when it was checked. 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.

normalize Ran
ran on a synthesized input
ispamm/hi2i/core/wing.py
code served (permissive licence) · get_code("5cb99cf6b4b4728e")
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 Tensor, ndim 4, dtype family float, shape (2, 4, 3, 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 torch

torch.manual_seed(0)

# The function normalize expects a 4D tensor (N, C, H, W)
# and applies min-max normalization across spatial dimensions
# Parameters: x (tensor), eps (float with default 1e-06)

x = torch.randn(2, 4, 3, 5)  # N=2, C=4, H=3, W=5

ARGS = {
    'x': x,
    'eps': 1e-06
}
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 Tensor, ndim 4, dtype family float, shape (2, 4, 3, 5) — shape (2, 4, 3, 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
preprocess Ran
ran on a synthesized input
ispamm/hi2i/core/wing.py
code served (permissive licence) · get_code("73f22265368bbe0a")
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 5, 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 torch
from collections import namedtuple
from munch import Munch

torch.manual_seed(0)

# The function preprocess takes a single argument x which should be a 4D tensor
# with shape (N, C, H, W) where:
# - N is batch size
# - C is 98 (number of facial landmark heatmaps)
# - H and W are spatial dimensions (height and width)

# Create a minimal valid input tensor
batch_size = 2
num_channels = 98
height = 256
width = 256

x = torch.randn(batch_size, num_channels, height, width)

ARGS = {
    'x': x
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
munch, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 5, 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
q_normalize Ran
ran; our contract draft was wrong, not the code
eleGAN23/QVAE/utils/quaternion_ops.py
pointer only (licence: MIT); the stored copy get_code serves, from XinxiiChen/AGREE, is pointer only (licence: NONE) · get_code("917981cbd6f167d2")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from XinxiiChen/AGREE, and GitHub detected no licence in that repository when it was checked. 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.

quaternion_init Ran
ran on a synthesized input
ispamm/hi2i/core/model.py
code served (permissive licence) · get_code("fd27db1d1f4c93de")
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 3, 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
from numpy.random import RandomState
from scipy.stats import chi

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

ARGS = {
    'in_features': 4,
    'out_features': 3,
    'rng': RandomState(42),
    'kernel_size': None,
    'criterion': 'glorot'
}
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 3, 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
shape Ran
ran on a synthesized input
ispamm/hi2i/htorch/quaternion.py
code served (permissive licence) · get_code("a7e5a2da84b86174")
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 Size. 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
import functools

torch.manual_seed(0)

HANDLEdFUNCTIONS = {}

def implements(torch_function):
    """Register a torch function override for ScalarTensor"""

    @functools.wraps(torch_function)
    def decorator(func):
        HANDLEdFUNCTIONS[torch_function] = func
        return func
    return decorator

@implements(torch.Tensor.shape.__get__)
def shape(self):
    """
    Gets standard shape
    """
    return torch.Tensor.shape.__get__(self.q)


# Create a minimal stand-in class that has a 'q' attribute which is a tensor
class ScalarTensor:
    def __init__(self):
        self.q = torch.randn(2, 4)


scalar_tensor_instance = ScalarTensor()

ARGS = {
    'self': scalar_tensor_instance
}
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 Size — a Size output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Size output the recorded ndim 1 and dtype family int 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
sigmoid Ran
ran on a synthesized input
ispamm/hi2i/core/utils.py
code served (permissive licence) · get_code("8b483042fac3893a")
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 (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

# The function sigmoid takes two parameters:
# - x: a numeric value or array
# - w: a weight parameter with default value 1

# Create a small numpy array for x
x = np.array([0.0, 1.0, -1.0, 2.0])

# Create a scalar weight parameter
w = 1.0

ARGS = {
    "x": x,
    "w": w
}
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 1, dtype family float, shape (4,) — shape (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
spectral_norm Ran
ran on a synthesized input
eleGAN23/QGAN/utils/Ops.py
code served (permissive licence) · get_code("855253f62eb9802c")
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 Linear. 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
import torch.nn as nn

torch.manual_seed(0)

# Create a simple linear module to apply spectral_norm to
module = nn.Linear(4, 3)

# The function applies torch.nn.utils.spectral_norm to a module
# with an optional name parameter (default is 'weight')
ARGS = {
    'module': module,
    'name': 'weight'
}
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 Linear — a Linear output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Linear 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
GetModel Not yet run
unverified (harvested, no recorded run)
eleGAN23/QGAN/GetModel.py
code served (permissive licence) · get_code("07ed353296f1a1fd")
How it ran

no invocation recorded.

apply_quaternion_gradient Not yet run
unverified (harvested, no recorded run)
ispamm/hi2i/htorch/utils.py
code served (permissive licence) · get_code("ae8f1566cec0edcf")
How it ran

no invocation recorded.

check_shapes Not yet run
unverified (harvested, no recorded run)
ispamm/hi2i/htorch/utils.py
code served (permissive licence) · get_code("21b315600a3a2e2b")
How it ran

no invocation recorded.

evaluate Not yet run
unverified (harvested, no recorded run)
elegan23/hypernets/sound-event-detection/train_baseline_task2.py
code served (permissive licence) · get_code("3e94054b554510ce")
How it ran

no invocation recorded.

inception_score Not yet run
unverified (harvested, no recorded run)
eleGAN23/QGAN/utils/inception_score.py
code served (permissive licence) · get_code("8def708af5968cc7")
How it ran

no invocation recorded.

seld_loss Not yet run
unverified (harvested, no recorded run)
elegan23/hypernets/sound-event-detection/train_baseline_task2.py
code served (permissive licence) · get_code("2c0bc31335e2f961")
How it ran

no invocation recorded.

Also present in other papers' code

Of 34 samples lifted from this paper's linked repositories, 34 carry a function identity and 0 do not. Of the 34 distinct functions with an identity, 10 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 2 the same month, 9 later) and 24 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
denormalize 2 other papers
2 later
2203.09301 (2022-03)
One-Shot Adaptation of GAN in Just One CLIP
get_alphas 2 other papers
1 the same month, 1 later
2110.11728 (2021-10)
BlendGAN: Implicitly GAN Blending for Arbitrary Stylized Face Generation
2203.09301 (2022-03)
One-Shot Adaptation of GAN in Just One CLIP
get_i 2 other papers
2 later
2204.05798 (2022-04)
Multi-View Hypercomplex Learning for Breast Cancer Screening
2606.23199 (2026-06)
Bridge the Gaps: Heterogeneous Attributed Graph Clustering via Quaternion Representation Learning
get_r 2 other papers
2 later
2204.05798 (2022-04)
Multi-View Hypercomplex Learning for Breast Cancer Screening
2606.23199 (2026-06)
Bridge the Gaps: Heterogeneous Attributed Graph Clustering via Quaternion Representation Learning
q_normalize 2 other papers
2 later
2204.05798 (2022-04)
Multi-View Hypercomplex Learning for Breast Cancer Screening
2606.23199 (2026-06)
Bridge the Gaps: Heterogeneous Attributed Graph Clustering via Quaternion Representation Learning
sigmoid 2 other papers
1 the same month, 1 later
2110.11728 (2021-10)
BlendGAN: Implicitly GAN Blending for Arbitrary Stylized Face Generation
2203.09301 (2022-03)
One-Shot Adaptation of GAN in Just One CLIP
adv_loss 1 other paper
1 dated earlier
2103.16835 (2021-03)
ReMix: Towards Image-to-Image Translation with Limited Data
frechet_distance 1 other paper
1 later
2203.09301 (2022-03)
One-Shot Adaptation of GAN in Just One CLIP
get_preds_fromhm 1 other paper
1 later
Park_LANIT_Language-Driven_Image-to-Image_Translation_for_Unlabeled_Data_CVPR_2023_paper (2023 (year only))
LANIT: Language-Driven Image-to-Image Translation for Unlabeled Data
moving_average_update 1 other paper
1 later
2606.18611 (2026-06)
QC-GAN: A Parameter-Efficient Quaternion Conformer GAN for High-Fidelity Speech Enhancement

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

Hypercomplex neural networks have proven to reduce the overall number of parameters while ensuring valuable performance by leveraging the properties of Clifford algebras. Recently, hypercomplex linear layers have been further improved by involving efficient parameterized Kronecker products. In this paper, we define the parameterization of hypercomplex convolutional layers and introduce the family of parameterized hypercomplex neural networks (PHNNs) that are lightweight and efficient large-scale models. Our method grasps the convolution rules and the filter organization directly from data without requiring a rigidly predefined domain structure to follow. PHNNs are flexible to operate in any user-defined or tuned domain, from 1D to $n$D regardless of whether the algebra rules are preset. Such a malleability allows processing multidimensional inputs in their natural domain without annexing further dimensions, as done, instead, in quaternion neural networks for 3D inputs like color images. As a result, the proposed family of PHNNs operates with $1/n$ free parameters as regards its analog in the real domain. We demonstrate the versatility of this approach to multiple domains of application by performing experiments on various image datasets as well as audio datasets in which our method outperforms real and quaternion-valued counterparts. Full code is available at: https://github.com/eleGAN23/HyperNets.

For agents

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

get_harvested_code_for_paper("2110.04176")
get_code_for_paper("2110.04176")
have("2110.04176")

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

curl https://syntology.ai/api/ran/2110.04176.json

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

[![Syntology run record](https://syntology.ai/api/ran/2110.04176.svg)](https://syntology.ai/paper/2110.04176)

Connect an agent — have() is free.