SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2306.04848 · ICML · 2024

Interpreting and Improving Diffusion Models from an Optimization Perspective

Frank Permenter, Chenyang Yuan

arXiv · PDF · Open in the Atlas

Code that ran

12 of 14 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; 1 honoured the contract we drafted, 0 violated it and 8 have no contract verdict; 3 failures of our instrument; 2 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/2306.04848.json · badge: /api/ran/2306.04848.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
ToyotaResearchInstitute/gradient-estimation-sampler canonical (classifier label)
classifier basis: census
9 ran · 2 no recorded run
toyotaresearchinstitute/gradient-estimation-sampler canonical (classifier label)
classifier basis: classifier_2025
3 ran · 0 no recorded run
FunctionStatusWhere it lives
GEScheduler Ran
ran on a synthesized input · constructed an object; no result was computed
toyotaresearchinstitute/gradient-estimation-sampler/gescheduler/scheduling_gradient_estimation.py
code served (permissive licence) · get_code("470ac5baf9cb29b5")
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, GEScheduler: 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 numpy as np
from dataclasses import dataclass

torch.manual_seed(0)

ARGS = {
    "num_train_timesteps": 1000,
    "beta_start": 0.0001,
    "beta_end": 0.02,
    "beta_schedule": "linear",
    "trained_betas": None,
    "clip_sample": True,
    "clip_sample_range": 1.0,
    "set_alpha_to_one": True,
    "prediction_type": "epsilon",
    "start": 14.6,
    "end": 0.02,
    "style": "DDIM",
    "gamma": torch.tensor(0.5, dtype=torch.float32),
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
diffusers, numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type GEScheduler — 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 GEScheduler 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
GESchedulerOutput Ran
ran on a synthesized input · constructed an object; no result was computed
toyotaresearchinstitute/gradient-estimation-sampler/gescheduler/scheduling_gradient_estimation.py
code served (permissive licence) · get_code("654c11faff8d06ac")
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, GESchedulerOutput: the run constructed the object and did not call it, so no result was computed. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
from dataclasses import dataclass
from typing import Optional

torch.manual_seed(0)

@dataclass
class GESchedulerOutput:
    """
    Output class for the scheduler's step function output.

    Args:
        prev_sample (`torch.FloatTensor` of shape `(batch_size, num_channels, height, width)` for images):
            Computed sample (x_{t-1}) of previous timestep. `prev_sample` should be used as next model input in the
            denoising loop.
        pred_original_sample (`torch.FloatTensor` of shape `(batch_size, num_channels, height, width)` for images):
            The predicted denoised sample (x_{0}) based on the model output from the current timestep.
            `pred_original_sample` can be used to preview progress or for guidance.
    """
    prev_sample: torch.FloatTensor
    pred_original_sample: Optional[torch.FloatTensor] = None

# Create tiny tensors with distinct dimensions
# batch_size=2, num_channels=4, height=3, width=5
prev_sample_tensor = torch.randn(2, 4, 3, 5)
pred_original_sample_tensor = torch.randn(2, 4, 3, 5)

ARGS = {
    "prev_sample": prev_sample_tensor,
    "pred_original_sample": pred_original_sample_tensor,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
diffusers, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type GESchedulerOutput — 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 GESchedulerOutput 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
Normalize Ran
ran; our contract draft was wrong, not the code
ToyotaResearchInstitute/gradient-estimation-sampler/experiments/ddim_model.py
pointer only (licence: MIT); the stored copy get_code serves, from Xingyu-Zheng/BiDM, is pointer only (licence: NONE) · get_code("c3a6b977022957cb")
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 Xingyu-Zheng/BiDM, 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.

batched_t Ran
ran on a synthesized input
ToyotaResearchInstitute/gradient-estimation-sampler/experiments/experiment_utils.py
code served (permissive licence) · get_code("ae9cd031d83e91e1")
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 1, dtype family int, shape (2,) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

ARGS = {
    "batchsize": 2,
    "t": 5,
    "device": torch.device("cpu")
}
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 1, dtype family int, shape (2,) — shape (2,) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
betas_for_alpha_bar Ran
ran, honoured the contract we drafted
toyotaresearchinstitute/gradient-estimation-sampler/gescheduler/scheduling_gradient_estimation.py
code served (permissive licence) · get_code("58ae788835842263")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned Tensor, ndim 1, dtype family float, shape (10,) 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 math

torch.manual_seed(0)

ARGS = {
    "num_diffusion_timesteps": 10,
    "max_beta": 0.999,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 1, dtype family float, shape (10,) — shape (10,) 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_timestep_embedding Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
ToyotaResearchInstitute/gradient-estimation-sampler/experiments/ddim_model.py
code served (permissive licence) · get_code("cb49209c125de1b4")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned 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
import math

torch.manual_seed(0)

# The function takes two parameters:
# 1. timesteps: a 1D tensor of timestep indices
# 2. embedding_dim: an integer specifying the embedding dimension

# Create a small 1D tensor of timesteps (batch size 2)
timesteps = torch.tensor([0, 5], dtype=torch.int64)

# Create an embedding dimension (use 4 as specified)
embedding_dim = 4

ARGS = {
    "timesteps": timesteps,
    "embedding_dim": embedding_dim,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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
nonlinearity Ran
ran; our contract draft was wrong, not the code
ToyotaResearchInstitute/gradient-estimation-sampler/experiments/ddim_model.py
code served (permissive licence) · get_code("3137073275f8c21a")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned 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 takes a single parameter 'x' which is used with:
# - multiplication operator (*)
# - torch.sigmoid() function
# This means x should be a tensor

x = torch.randn(2, 4)

ARGS = {
    "x": x
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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
norm Ran
ran on a synthesized input
ToyotaResearchInstitute/gradient-estimation-sampler/experiments/experiment_utils.py
code served (permissive licence) · get_code("a7e279fafe623d56")
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 1, dtype family float, shape (2,) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# The function norm() takes a single parameter x and computes vector norms
# It expects x to be a tensor with at least 4 dimensions (since it reduces over dims 1,2,3)
# We'll create a small 4D tensor: (batch=2, dim1=4, dim2=3, dim3=5)
x = torch.randn(2, 4, 3, 5)

ARGS = {
    'x': x
}
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 Tensor, ndim 1, dtype family float, shape (2,) — shape (2,) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
projections Ran
ran on a synthesized input
ToyotaResearchInstitute/gradient-estimation-sampler/experiments/ideal_denoiser.py
code served (permissive licence) · get_code("3e7d0174c03ef683")
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)

# Create a simple dataset class that supports indexing
class SimpleDataset:
    def __init__(self, data):
        self.data = data
    
    def __getitem__(self, idx):
        return self.data[idx]
    
    def __len__(self):
        return len(self.data)

# Create sample data: batch_size=2, feature_dim=4
data_samples = [
    (torch.randn(4), 0),
    (torch.randn(4), 1),
    (torch.randn(4), 0),
    (torch.randn(4), 1),
]

dataset = SimpleDataset(data_samples)

# Create a simple data loader (list of batches)
loader = [
    (torch.randn(2, 4), torch.tensor([0, 1])),
    (torch.randn(2, 4), torch.tensor([1, 0])),
]

# Create input x with shape (2, 4) to match loader batch dimensions
x = torch.randn(2, 4)

ARGS = {
    'dataset': dataset,
    'loader': loader,
    '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 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
show Ran
ran on a synthesized input
ToyotaResearchInstitute/gradient-estimation-sampler/experiments/experiment_utils.py
code served (permissive licence) · get_code("4a2eec11353e2317")
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 Image. 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 torchvision.transforms import functional as TF
from torchvision.utils import make_grid, save_image

torch.manual_seed(0)

# The function expects:
# - x: a tensor that can be passed to make_grid
# - nrow: int for number of rows
# - padding: int for padding

# make_grid expects a 4D tensor (batch, channels, height, width)
# The function applies .add(1).div(2).clamp(0, 1) which are tensor operations
# to_pil_image expects a 3D tensor (channels, height, width)

x = torch.randn(2, 3, 4, 5)  # batch=2, channels=3, height=4, width=5

ARGS = {
    'x': x,
    'nrow': 4,
    'padding': 2
}
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 Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 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
sq_norm Ran
ran on a synthesized input
ToyotaResearchInstitute/gradient-estimation-sampler/experiments/ideal_denoiser.py
code served (permissive licence) · get_code("e4f8a62a2e20ec8c")
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, 3) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# M should be a 2D tensor where dim=1 gives the second dimension
# The function computes norm along dim=1, so we need a batch dimension
# k is used in repeat(1, k) to repeat the result k times
M = torch.randn(2, 4)  # batch size 2, feature dim 4
k = 3

ARGS = {
    "M": M,
    "k": k
}
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, 3) — shape (2, 3) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
tensor_to_pil Ran
ran on a synthesized input
ToyotaResearchInstitute/gradient-estimation-sampler/experiments/stable_diffusion_exp.py
code served (permissive licence) · get_code("d4f6bea07d7f48a0")
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 Image. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
from PIL import Image

torch.manual_seed(0)

# The function expects a tensor with shape (C, H, W) where:
# - C is channels (3 for RGB)
# - H is height
# - W is width
# The tensor values should be in range [0, 1] to produce valid uint8 after scaling

img_tensor = torch.rand(3, 4, 5)  # (channels=3, height=4, width=5)

ARGS = {
    "img": img_tensor
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Image — a Image output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a Image output the recorded ndim 3 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
get_experiment_celeba Not yet run
unverified (harvested, no recorded run)
ToyotaResearchInstitute/gradient-estimation-sampler/experiments/run_experiments.py
code served (permissive licence) · get_code("2ba5b12cc97477f9")
How it ran

no invocation recorded.

get_experiment_cifar10 Not yet run
unverified (harvested, no recorded run)
ToyotaResearchInstitute/gradient-estimation-sampler/experiments/run_experiments.py
code served (permissive licence) · get_code("69f576a7867d0e84")
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, 4 are also present in other papers whose code we hold (3 with at least one paper dated earlier, 3 the same month, 4 later) and 10 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
nonlinearity 156 other papers
36 dated earlier, 4 the same month, 115 later, 1 same year, order not determinable · 25 only as an assembly brick
2104.07652 (2021-04)
Geometry-Free View Synthesis: Transformers and no 3D Priors
2106.00132 (2021-06)
On Fast Sampling of Diffusion Probabilistic Models
2106.06819 (2021-06)
D2C: Diffusion-Denoising Models for Few-shot Conditional Generation
and 153 more not named here
get_timestep_embedding 91 other papers
29 dated earlier, 2 the same month, 59 later, 1 same year, order not determinable · 7 only as an assembly brick
2104.07652 (2021-04)
Geometry-Free View Synthesis: Transformers and no 3D Priors
2106.00132 (2021-06)
On Fast Sampling of Diffusion Probabilistic Models
2106.06819 (2021-06)
D2C: Diffusion-Denoising Models for Few-shot Conditional Generation
and 88 more not named here
Normalize 90 other papers
30 dated earlier, 2 the same month, 58 later · 32 only as an assembly brick
2104.07652 (2021-04)
Geometry-Free View Synthesis: Transformers and no 3D Priors
2106.00132 (2021-06)
On Fast Sampling of Diffusion Probabilistic Models
2106.06819 (2021-06)
D2C: Diffusion-Denoising Models for Few-shot Conditional Generation
and 87 more not named here
betas_for_alpha_bar
brick of this paper's assembly: helper
3 other papers
3 later
2308.13712 (2023-08)
Residual Denoising Diffusion Models
2312.02813 (2023-12)
BIVDiff: A Training-Free Framework for General-Purpose Video Synthesis via Bridging Image and Video Diffusion Models
2403.11157 (2024-03)
Selective Hourglass Mapping for Universal Image Restoration Based on Diffusion Model

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

Denoising is intuitively related to projection. Indeed, under the manifold hypothesis, adding random noise is approximately equivalent to orthogonal perturbation. Hence, learning to denoise is approximately learning to project. In this paper, we use this observation to interpret denoising diffusion models as approximate gradient descent applied to the Euclidean distance function. We then provide straightforward convergence analysis of the DDIM sampler under simple assumptions on the projection error of the denoiser. Finally, we propose a new gradient-estimation sampler, generalizing DDIM using insights from our theoretical results. In as few as 5-10 function evaluations, our sampler achieves state-of-the-art FID scores on pretrained CIFAR-10 and CelebA models and can generate high quality samples on latent diffusion models.

For agents

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

get_harvested_code_for_paper("2306.04848")
get_code_for_paper("2306.04848")
have("2306.04848")

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

curl https://syntology.ai/api/ran/2306.04848.json

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

[![Syntology run record](https://syntology.ai/api/ran/2306.04848.svg)](https://syntology.ai/paper/2306.04848)

Connect an agent — have() is free.