SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2012.15000 · ICLR · 2020

DeepSphere: a graph-based spherical CNN

Michaël Defferrard, Martino Milani, Frédérick Gusset, Nathanaël Perraudin

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 (2 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 7 have no contract verdict; 2 failures of our instrument; 4 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/2012.15000.json · badge: /api/ran/2012.15000.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
deepsphere/deepsphere-cosmo-tf1 canonical (classifier label)
classifier basis: classifier_2025
0 ran · 4 no recorded run
deepsphere/deepsphere-pytorch port (classifier label)
classifier basis: classifier_2025
3 ran · 0 no recorded run
smsharma/neural-global-astrometry community
Papers with Code archive
3 ran · 0 no recorded run
deepsphere/deepsphere-weather application (classifier label)
classifier basis: classifier_2025
2 ran · 0 no recorded run
smsharma/fermi-gce-flows community
Papers with Code archive
1 ran · 0 no recorded run
FunctionStatusWhere it lives
ChebConv Ran
ran on a synthesized input · constructed an object; no result was computed
deepsphere/deepsphere-pytorch/deepsphere/layers/chebyshev.py
code served (permissive licence) · get_code("faa9c3590068c78f")
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, ChebConv: 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
import math

torch.manual_seed(0)

# Parameters for ChebConv
in_channels = 3
out_channels = 5
kernel_size = 4
bias = True
conv = None  # Will use the cheb_conv function defined in the module

ARGS = {
    "in_channels": in_channels,
    "out_channels": out_channels,
    "kernel_size": kernel_size,
    "bias": bias,
    "conv": conv
}
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 ChebConv — 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 ChebConv 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
ChebConv Ran
ran on a synthesized input
smsharma/neural-global-astrometry/models/layers.py
pointer only (licence: AGPL-3.0) · get_code("516e9a5f45631fe6")
How it ran

The body of this sample is withheld: AGPL-3.0, as GitHub detects it on this paper's repository, smsharma/neural-global-astrometry, is not in the permissive set Syntology serves inline. 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.

ConvCheb Ran
ran on a synthesized input
deepsphere/deepsphere-weather/modules/layers.py
code served (permissive licence) · get_code("3c83b16df98a066f")
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 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 torch
import math

def conv_cheb(laplacian, inputs, weight):
    """Chebyshev convolution."""
    B, V, Fin1 = inputs.shape
    Fin, K, Fout = weight.shape
    if not Fin1 == Fin:
        raise ValueError('Input tensor shape does not match the expected shape: \n' + '- Input tensor shape :{} \n'.format(Fin1) + '- Expected tensor shape :{} \n'.format(Fin))
    x0 = inputs.permute(1, 2, 0).contiguous()
    x0 = x0.view([V, Fin * B])
    x = x0.unsqueeze(0)
    if K > 1:
        x1 = torch.sparse.mm(laplacian, x0)
        x = torch.cat((x, x1.unsqueeze(0)), 0)
    for _ in range(2, K):
        x2 = 2 * torch.sparse.mm(laplacian, x1) - x0
        x = torch.cat((x, x2.unsqueeze(0)), 0)
        x0, x1 = (x1, x2)
    x = x.view([K, V, Fin, B])
    x = x.permute(3, 1, 2, 0).contiguous()
    x = x.view([B * V, Fin * K])
    weight = weight.view(Fin * K, Fout)
    x = x.matmul(weight)
    x = x.view([B, V, Fout])
    return x

class ConvCheb(torch.nn.Module):
    """Graph Convolutional Layer based on Chebyshev polynomials."""

    def __init__(self, in_channels, out_channels, kernel_size, laplacian, bias=True, conv=conv_cheb, **kwargs):
        super().__init__()
        self.in_channels = in_channels
        self.out_channels = out_channels
        self.kernel_size = kernel_size
        self._conv = conv
        self.register_buffer('laplacian', laplacian)
        shape = (in_channels, kernel_size, out_channels)
        self.weight = torch.nn.Parameter(torch.Tensor(*shape))
        if bias:
            self.bias = torch.nn.Parameter(torch.Tensor(out_channels))
        else:
            self.register_parameter('bias', None)
        self.reset_parameters()

    def reset_parameters(self, activation='relu', fan='in', distribution='normal'):
        """Reset weight and bias."""
        if activation == 'relu':
            gain = math.sqrt(2.0)
        else:
            gain = 1.0
        
        if fan == 'in':
            fan_value = self.in_channels * self.kernel_size
        elif fan == 'out':
            fan_value = self.out_channels * self.kernel_size
        else:
            fan_value = (self.in_channels + self.out_channels) * self.kernel_size / 2.0
        
        std = gain / math.sqrt(fan_value)
        
        if distribution == 'normal':
            self.weight.data.normal_(0, std)
        else:
            bound = math.sqrt(3.0) * std
            self.weight.data.uniform_(-bound, bound)
        
        if self.bias is not None:
            self.bias.data.fill_(0)

    def forward(self, inputs):
        """Forward pass."""
        x = self._conv(self.laplacian, inputs, self.weight)
        if self.bias is not None:
            x = x + self.bias
        return x


# Create a small sparse Laplacian matrix for a graph with 5 vertices
V = 5
indices = torch.LongTensor([[0, 1, 1, 2, 2, 3, 3, 4, 4, 0],
                             [1, 0, 2, 1, 3, 2, 4, 3, 0, 4]])
values = torch.FloatTensor([-1.0, -1.0, -1.0, -1.0, -1.0, -1.0, -1.0, -1.0, -1.0, -1.0])
laplacian = torch.sparse.FloatTensor(indices, values, torch.Size([V, V]))

INIT_ARGS = {
    'in_channels': 3,
    'out_channels': 4,
    'kernel_size': 2,
    'laplacian': laplacian,
    'bias': True,
    'conv': conv_cheb,
}

ARGS = {
    'inputs': torch.randn(2, 5, 3),
}
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 3, dtype family float — no output shape is stated: observe_signature.py could not observe it (harness:RuntimeError)
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)
SphericalChebBNPool Ran
ran on a synthesized input
smsharma/neural-global-astrometry/models/layers.py
pointer only (licence: AGPL-3.0) · get_code("e7b4a817174e8ab8")
How it ran

The body of this sample is withheld: AGPL-3.0, as GitHub detects it on this paper's repository, smsharma/neural-global-astrometry, is not in the permissive set Syntology serves inline. 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.

SphericalChebConv Ran
ran on a synthesized input · constructed an object; no result was computed
deepsphere/deepsphere-pytorch/deepsphere/layers/chebyshev.py
code served (permissive licence) · get_code("68383e49022b1b3d")
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, SphericalChebConv: 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
import torch.nn as nn
import math

torch.manual_seed(0)

# Create a small sparse Laplacian matrix
# For a graph with V=4 vertices
V = 4
# Create a simple sparse adjacency-like structure
indices = torch.tensor([[0, 1, 2, 3, 0, 1], [1, 0, 3, 2, 0, 1]], dtype=torch.long)
values = torch.tensor([0.5, 0.5, 0.5, 0.5, -1.0, -1.0], dtype=torch.float)
laplacian = torch.sparse_coo_tensor(indices, values, (V, V))

# Input tensor: [batch, vertices, in_channels]
batch_size = 2
in_channels = 3
inputs = torch.randn(batch_size, V, in_channels)

# Parameters
out_channels = 5
kernel_size = 3

ARGS = {
    'in_channels': in_channels,
    'out_channels': out_channels,
    'lap': laplacian,
    'kernel_size': kernel_size,
}
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 SphericalChebConv — 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 SphericalChebConv 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
SphericalChebConv Ran
ran on a synthesized input
smsharma/neural-global-astrometry/models/layers.py
pointer only (licence: AGPL-3.0) · get_code("108a6d8353096846")
How it ran

The body of this sample is withheld: AGPL-3.0, as GitHub detects it on this paper's repository, smsharma/neural-global-astrometry, is not in the permissive set Syntology serves inline. 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.

SphericalChebConv Ran
ran on a synthesized input
smsharma/fermi-gce-flows/models/chebyshev.py
pointer only (licence: NONE) · get_code("0a93841faa89f61a")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, smsharma/fermi-gce-flows 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.

cheb_conv Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
deepsphere/deepsphere-pytorch/deepsphere/layers/chebyshev.py
code served (permissive licence) · get_code("e3a899ff8c3b7586")
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, 6) 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)

# Small dimensions for distinct sizes
B = 2      # batch size
V = 4      # number of vertices
Fin = 3    # input features
K = 5      # Chebyshev polynomial order
Fout = 6   # output features

# Create a sparse Laplacian matrix (V x V)
# Use a simple sparse matrix structure for the sphere/graph
indices = torch.LongTensor([
    [0, 1, 2, 3, 0, 1, 2, 3],
    [0, 1, 2, 3, 1, 0, 3, 2]
])
values = torch.FloatTensor([-2.0, -2.0, -2.0, -2.0, 1.0, 1.0, 1.0, 1.0])
laplacian = torch.sparse_coo_tensor(indices, values, (V, V), dtype=torch.float32)

# Create inputs tensor (B, V, Fin)
inputs = torch.randn(B, V, Fin)

# Create weight tensor (K, Fin, Fout)
weight = torch.randn(K, Fin, Fout)

ARGS = {
    "laplacian": laplacian,
    "inputs": inputs,
    "weight": 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 Tensor, ndim 3, dtype family float, shape (2, 4, 6) — shape (2, 4, 6) 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
conv_cheb Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
deepsphere/deepsphere-weather/modules/layers.py
code served (permissive licence) · get_code("428398b69dc7689d")
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, 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)

# Parameters for conv_cheb
B = 2  # batch size
V = 4  # number of vertices
Fin = 3  # input features
K = 2  # Chebyshev polynomial order
Fout = 5  # output features

# Create a sparse laplacian matrix (V x V)
# Using a simple sparse identity-like matrix for testing
indices = torch.tensor([[0, 1, 2, 3], [0, 1, 2, 3]], dtype=torch.long)
values = torch.tensor([-2.0, -2.0, -2.0, -2.0], dtype=torch.float32)
laplacian = torch.sparse_coo_tensor(indices, values, (V, V), dtype=torch.float32)

# Create input tensor (B, V, Fin)
inputs = torch.randn(B, V, Fin, dtype=torch.float32)

# Create weight tensor (Fin, K, Fout)
weight = torch.randn(Fin, K, Fout, dtype=torch.float32)

ARGS = {
    "laplacian": laplacian,
    "inputs": inputs,
    "weight": 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 Tensor, ndim 3, dtype family float, shape (2, 4, 5) — shape (2, 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
LoadableGenerator Not yet run
unverified (harvested, no recorded run)
deepsphere/deepsphere-cosmo-tf1/deepsphere/models.py
code served (permissive licence) · get_code("a12561557653c9c0")
How it ran

no invocation recorded.

base_model Not yet run
unverified (harvested, no recorded run)
deepsphere/deepsphere-cosmo-tf1/deepsphere/models.py
code served (permissive licence) · get_code("bc9700d8a529944a")
How it ran

no invocation recorded.

cgcnn Not yet run
unverified (harvested, no recorded run)
deepsphere/deepsphere-cosmo-tf1/deepsphere/models.py
code served (permissive licence) · get_code("67a2d19c21df34ca")
How it ran

no invocation recorded.

deepsphere Not yet run
unverified (harvested, no recorded run)
deepsphere/deepsphere-cosmo-tf1/deepsphere/models.py
code served (permissive licence) · get_code("630e219d155f863b")
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 11 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 11 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-21.

Repositories linked to this paper

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

Abstract

Designing a convolution for a spherical neural network requires a delicate tradeoff between efficiency and rotation equivariance. DeepSphere, a method based on a graph representation of the sampled sphere, strikes a controllable balance between these two desiderata. This contribution is twofold. First, we study both theoretically and empirically how equivariance is affected by the underlying graph with respect to the number of vertices and neighbors. Second, we evaluate DeepSphere on relevant problems. Experiments show state-of-the-art performance and demonstrates the efficiency and flexibility of this formulation. Perhaps surprisingly, comparison with previous work suggests that anisotropic filters might be an unnecessary price to pay. Our code is available at https: //github.com/deepsphere.

For agents

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

get_harvested_code_for_paper("2012.15000")
get_code_for_paper("2012.15000")
have("2012.15000")

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

curl https://syntology.ai/api/ran/2012.15000.json

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

[![Syntology run record](https://syntology.ai/api/ran/2012.15000.svg)](https://syntology.ai/paper/2012.15000)

Connect an agent — have() is free.