SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2007.02798 · ICLR · 2021

Gradient Origin Networks

Sam Bond-Taylor, Chris Willcocks

arXiv · PDF · Open in the Atlas

Code that ran

11 of 14 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (4 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 0 violated it and 10 have no contract verdict; 0 failures of our instrument; 3 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/2007.02798.json · badge: /api/ran/2007.02798.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
cwkx/GON canonical (classifier label)
classifier basis: census
4 ran · 0 no recorded run
cwkx/gon canonical (classifier label)
classifier basis: classifier_2025
1 ran · 0 no recorded run
kklemon/gon-pytorch community
Papers with Code archive
3 ran · 2 no recorded run
titu1994/tf_GON community (classifier label)
classifier basis not recorded on the link
2 ran · 1 no recorded run
BariscanBozkurt/Gradient-Origin-Networks-Julia community
Papers with Code archive
1 ran · 0 no recorded run
FunctionStatusWhere it lives
CoordinateEncoding Ran
ran on a synthesized input · constructed an object; no result was computed
kklemon/gon-pytorch/gon_pytorch/modules.py
code served (permissive licence) · get_code("aaa4929536e5ac6b")
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, CoordinateEncoding: 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

torch.manual_seed(0)

# Create a projection matrix
# in_dim = 4, out_dim = 3, so proj_matrix shape is (4, 3)
proj_matrix = torch.randn(4, 3)

ARGS = {
    'proj_matrix': proj_matrix,
    'is_trainable': False
}
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 CoordinateEncoding — 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 CoordinateEncoding 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
GON Ran
ran on a synthesized input · constructed an object; no result was computed
kklemon/gon-pytorch/gon_pytorch/modules.py
code served (permissive licence) · get_code("ec7d9ee49751035d")
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, GON: 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
from math import pi

torch.manual_seed(0)

class SimpleBlockFactory:
    def __call__(self, in_f, out_f, is_first=False, is_last=False):
        return nn.Linear(in_f, out_f)

# Build proj_matrix for CoordinateEncoding
# Input dim: 2, output dim will be 2*2=4 after sin/cos
proj_matrix = torch.randn(2, 2)

# Create CoordinateEncoding
pos_encoder = nn.Module.__new__(type('CoordinateEncoding', (nn.Module,), {
    '__init__': lambda self, proj_matrix, is_trainable=False: (
        nn.Module.__init__(self),
        (self.register_parameter('proj_matrix', nn.Parameter(proj_matrix)) if is_trainable 
         else self.register_buffer('proj_matrix', proj_matrix)),
        setattr(self, 'in_dim', proj_matrix.size(0)),
        setattr(self, 'out_dim', proj_matrix.size(1) * 2)
    )[-1],
    'forward': lambda self, x: (
        torch.cat([
            torch.sin(2 * pi * (x.reshape(-1, x.shape[-1]) @ self.proj_matrix).view(*x.shape[:-1], -1)),
            torch.cos(2 * pi * (x.reshape(-1, x.shape[-1]) @ self.proj_matrix).view(*x.shape[:-1], -1))
        ], dim=-1)
    )
}))
pos_encoder.__init__(proj_matrix, is_trainable=False)

# Create decoder
latent_dim = 3
out_dim = 5
hidden_dim = 8
num_layers = 2
block_factory = SimpleBlockFactory()

decoder = nn.Module.__new__(type('ImplicitDecoder', (nn.Module,), {
    '__init__': lambda self, latent_dim, out_dim, hidden_dim, num_layers, block_factory, pos_encoder=None, modulation=False, dropout=0.0, final_activation=torch.sigmoid: (
        nn.Module.__init__(self),
        setattr(self, 'pos_encoder', pos_encoder),
        setattr(self, 'latent_dim', latent_dim),
        setattr(self, 'mod_network', None),
        setattr(self, 'net', nn.Sequential(
            nn.Linear(pos_encoder.out_dim + latent_dim, hidden_dim),
            nn.ReLU(),
            nn.Linear(hidden_dim, out_dim)
        ))
    )[-1],
    'forward': lambda self, input, latent: self.net(
        torch.cat([
            latent.view(latent.shape[0], 1, 1, -1).expand(input.shape[0], *input.shape[1:-1], latent.shape[-1]),
            self.pos_encoder(input)
        ], dim=-1)
    ) if self.pos_encoder else input
}))
decoder.__init__(latent_dim, out_dim, hidden_dim, num_layers, block_factory, pos_encoder, modulation=False, dropout=0.0, final_activation=torch.sigmoid)

# Create GON arguments
ARGS = {
    'decoder': decoder,
    'latent_updates': 1,
    'learn_origin': False
}
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 GON — 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 GON 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
Generator Ran
ran on a synthesized input
cwkx/gon/GON.py
code served (permissive licence) · get_code("ce22901dbfe25134")
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 4, 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

nc = 1
ngf = 16
nz = 32

class Generator(nn.Module):

    def __init__(self):
        super(Generator, self).__init__()
        self.main = nn.Sequential(nn.ConvTranspose2d(nz, ngf * 4, 4, 1, 0, bias=True), nn.BatchNorm2d(ngf * 4), nn.ELU(), nn.ConvTranspose2d(ngf * 4, ngf * 2, 4, 2, 1, bias=True), nn.BatchNorm2d(ngf * 2), nn.ELU(), nn.ConvTranspose2d(ngf * 2, ngf, 4, 2, 1, bias=True), nn.BatchNorm2d(ngf), nn.ELU(), nn.ConvTranspose2d(ngf, nc, 4, 2, 1, bias=True), nn.Sigmoid())

    def forward(self, input):
        return self.main(input)


INIT_ARGS = {}

ARGS = {
    'input': torch.randn(2, nz, 1, 1)
}

if __name__ == '__main__':
    gen = Generator(**INIT_ARGS)
    output = gen(**ARGS)
    print(f"Output shape: {output.shape}")
    print(f"Output min: {output.min()}, max: {output.max()}")
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 4, dtype family float — no output shape is stated: observe_signature.py observed this class with its constructor sizes canonicalised where it could, which the node does not record, so its shape is not stated as this setup block's
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)
Generator Ran
ran on a synthesized input · constructed an object; no result was computed
BariscanBozkurt/Gradient-Origin-Networks-Julia/Pytorch_Debug_Files/Variational_GON_Train_MNIST.py
pointer only (licence: NONE) · get_code("47d79135f97a756f")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, BariscanBozkurt/Gradient-Origin-Networks-Julia 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.

MLP Ran
ran on a synthesized input · constructed an object; no result was computed
kklemon/gon-pytorch/gon_pytorch/modules.py
code served (permissive licence) · get_code("3d41ac95604b3889")
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, MLP: 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
from typing import Optional, Callable

torch.manual_seed(0)

# Define a concrete implementation of BaseBlockFactory
class SimpleBlockFactory(nn.Module.__class__):
    def __call__(self, in_f, out_f, is_first=False, is_last=False):
        return nn.Linear(in_f, out_f)

# Create a simple block factory instance
class ConcreteBlockFactory:
    def __call__(self, in_f, out_f, is_first=False, is_last=False):
        return nn.Linear(in_f, out_f)

# Define a simple final activation function
def simple_activation(x):
    return torch.relu(x)

# Construct ARGS dictionary
ARGS = {
    'in_dim': 4,
    'out_dim': 3,
    'hidden_dim': 8,
    'num_layers': 2,
    'block_factory': ConcreteBlockFactory(),
    'dropout': 0.1,
    'final_activation': simple_activation,
}
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 MLP — 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 MLP 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
get_mgrid Ran
ran, honoured the contract we drafted
cwkx/GON/Implicit-GON.py
code served (permissive licence) · get_code("6c563c92e8bf87cf")
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, 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)

ARGS = {
    "sidelen": 4,
    "dim": 2,
}
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 (16, 2) — shape (16, 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_mgrid Ran
ran on a synthesized input
titu1994/tf_GON/utils.py
code served (permissive licence) · get_code("6a41e809bc39a97f")
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 EagerTensor, ndim 2, dtype family float, shape (25, 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 tensorflow as tf

ARGS = {
    "sidelen": 5,
    "dim": 2,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, ndim 2, dtype family float, shape (25, 2) — shape (25, 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
gon_model Ran
ran on a synthesized input
cwkx/GON/Implicit-GON.py
code served (permissive licence) · get_code("35cc31140fb1f2c7")
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 Sequential. 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 numpy as np

torch.manual_seed(0)

# The gon_model function takes a single parameter: dimensions
# Looking at the function body:
# - dimensions is indexed with [0], [1], [2:-2], [2:-1], [-2], [-1]
# - It's used in zip(dimensions[1:-2], dimensions[2:-1])
# - It needs at least 4 elements to work properly (first_layer uses [0] and [1],
#   final_layer uses [-2] and [-1], and we need at least one intermediate layer)

# Create a dimensions list with distinct sizes
dimensions = [4, 8, 16, 32, 7]

ARGS = {
    "dimensions": dimensions
}
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 Sequential — a Sequential output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Sequential output the recorded ndim 1 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
slerp Ran
ran on a synthesized input
cwkx/GON/Implicit-GON.py
code served (permissive licence) · get_code("f685a32ca62b8db0")
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)

# Create two random normalized vectors with batch size 2 and dimension 4
a = torch.randn(2, 4)
b = torch.randn(2, 4)

# Create interpolation parameter t
t = 0.5

ARGS = {
    "a": a,
    "b": b,
    "t": t,
}
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
slerp Ran
ran on a synthesized input
titu1994/tf_GON/utils.py
code served (permissive licence) · get_code("def3c6042ace1a88")
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 EagerTensor, 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 tensorflow as tf
import numpy as np

# Create small batch of normalized vectors
# batch_size=2, embedding_dim=4
a_data = np.array([[1.0, 0.0, 0.0, 0.0], [0.0, 1.0, 0.0, 0.0]], dtype=np.float32)
b_data = np.array([[0.0, 0.0, 1.0, 0.0], [0.0, 0.0, 0.0, 1.0]], dtype=np.float32)

a = tf.constant(a_data)
b = tf.constant(b_data)
t = 0.5

ARGS = {
    "a": a,
    "b": b,
    "t": t
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, tensorflow — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type EagerTensor, 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
vae_loss Ran
ran on a synthesized input
cwkx/GON/Variational-GON.py
code served (permissive licence) · get_code("914a6c717f54ca6c")
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)

nc = 1

# The function expects:
# p: tensor that can be viewed as (-1, 32 * 32 * nc) = (-1, 1024)
# x: tensor that can be viewed as (-1, 32 * 32 * nc) = (-1, 1024)
# mu: tensor with shape (batch_size, latent_dim) for sum over dim=1
# logvar: tensor with shape (batch_size, latent_dim) for sum over dim=1
# weight: scalar (default 0.0)

batch_size = 2
latent_dim = 4
image_flat_size = 32 * 32 * nc  # 1024

# Create p as reconstructed output (batch_size, 1024) with values in valid range for BCE
p = torch.sigmoid(torch.randn(batch_size, image_flat_size))

# Create x as target (batch_size, 1024) with values in [0, 1] for BCE
x = torch.rand(batch_size, image_flat_size)

# Create mu (batch_size, latent_dim)
mu = torch.randn(batch_size, latent_dim)

# Create logvar (batch_size, latent_dim)
logvar = torch.randn(batch_size, latent_dim)

# weight is optional with default 0.0, provide a scalar
weight = 1.0

ARGS = {
    'p': p,
    'x': x,
    'mu': mu,
    'logvar': logvar,
    '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 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
BaseBlockFactory Not yet run
unverified (harvested, no recorded run)
kklemon/gon-pytorch/gon_pytorch/modules.py
code served (permissive licence) · get_code("7ba09778f338ca24")
How it ran

no invocation recorded.

ImplicitDecoder Not yet run
unverified (harvested, no recorded run)
kklemon/gon-pytorch/gon_pytorch/modules.py
code served (permissive licence) · get_code("3bbc204efec08fa6")
How it ran

no invocation recorded.

slerp_batch Not yet run
unverified (harvested, no recorded run)
titu1994/tf_GON/utils.py
code served (permissive licence) · get_code("1ee0daf20ef5831a")
How it ran

no invocation recorded.

Also present in other papers' code

Of 14 samples lifted from this paper's linked repositories, 14 carry a function identity and 0 do not. Of the 14 distinct functions with an identity, 1 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 1 the same month, 1 later) and 13 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
get_mgrid 3 other papers
1 the same month, 2 later
2007.09601 (2020-07)
Hypersolvers: Toward Fast Continuous-Depth Models
2406.02529 (2024-06)
ReLUs Are Sufficient for Learning Implicit Neural Representations
2406.03873 (2024-06)
Quantum Implicit Neural Representations

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

This paper proposes a new type of generative model that is able to quickly learn a latent representation without an encoder. This is achieved using empirical Bayes to calculate the expectation of the posterior, which is implemented by initialising a latent vector with zeros, then using the gradient of the log-likelihood of the data with respect to this zero vector as new latent points. The approach has similar characteristics to autoencoders, but with a simpler architecture, and is demonstrated in a variational autoencoder equivalent that permits sampling. This also allows implicit representation networks to learn a space of implicit functions without requiring a hypernetwork, retaining their representation advantages across datasets. The experiments show that the proposed method converges faster, with significantly lower reconstruction error than autoencoders, while requiring half the parameters.

For agents

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

get_harvested_code_for_paper("2007.02798")
get_code_for_paper("2007.02798")
have("2007.02798")

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

curl https://syntology.ai/api/ran/2007.02798.json

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

[![Syntology run record](https://syntology.ai/api/ran/2007.02798.svg)](https://syntology.ai/paper/2007.02798)

Connect an agent — have() is free.