SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2201.11945 · 2022

Learning Proximal Operators to Discover Multiple Optima

arXiv · PDF · Open in the Atlas

Code that ran

11 of 11 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; 0 honoured the contract we drafted, 0 violated it and 11 have no contract verdict; 0 failures of our instrument; 0 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/2201.11945.json · badge: /api/ran/2201.11945.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
lingxiaoli94/pol canonical (classifier label)
classifier basis not recorded on the link
11 ran · 0 no recorded run
FunctionStatusWhere it lives
aggregate Ran
ran on a synthesized input
lingxiaoli94/pol/pol/utils/loss_aggregation.py
code served (permissive licence) · get_code("063a0fdf435680a0")
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 a small tensor L with distinct dimensions
# Using shape (2, 3, 4) where dim 0 is batch, dim 1 is feature, dim 2 is another dimension
L = torch.randn(2, 3, 4)

# Test with 'mean' mode
desc_mean = {'mode': 'mean'}

# Test with 'max' mode
desc_max = {'mode': 'max'}

# Test with 'lse' mode
desc_lse = {
    'mode': 'lse',
    'lse_theta': 0.5
}

# We'll construct ARGS for the 'mean' mode as the primary test case
ARGS = {
    'L': L,
    'dim': 1,
    'desc': desc_mean
}
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
collate_padding_fn Ran
ran on a synthesized input
lingxiaoli94/pol/pol/datasets/objdetect.py
code served (permissive licence) · get_code("9d1ae5336498c0a1")
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 dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Create sample data items with images and boxes
data_item_1 = {
    'img': torch.randn(3, 4, 5),  # C=3, H=4, W=5
    'boxes': torch.tensor([[1.0, 2.0, 3.0, 4.0], [5.0, 6.0, 7.0, 8.0]]),  # 2 boxes with 4 coords
}

data_item_2 = {
    'img': torch.randn(3, 6, 7),  # C=3, H=6, W=7 (different sizes)
    'boxes': torch.tensor([[0.5, 1.5, 2.5, 3.5]]),  # 1 box
}

data_item_3 = {
    'img': torch.randn(3, 5, 6),  # C=3, H=5, W=6
    'boxes': torch.tensor([]),  # Empty boxes
}

# Reshape empty boxes to have correct shape
data_item_3['boxes'] = data_item_3['boxes'].reshape(0, 4)

data_list = [data_item_1, data_item_2, data_item_3]

ARGS = {
    'data_list': data_list
}
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 dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict 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
compute_model_size Ran
ran on a synthesized input
lingxiaoli94/pol/pol/utils/model_size.py
code served (permissive licence) · get_code("14b4684487d5b6e6")
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 float, ndim 0, dtype family float, 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 SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear1 = nn.Linear(4, 3)
        self.linear2 = nn.Linear(3, 2)
        self.buffer_tensor = torch.randn(5, 6)
        self.register_buffer('my_buffer', self.buffer_tensor)
    
    def forward(self, x):
        x = self.linear1(x)
        x = self.linear2(x)
        return x

model = SimpleModel()

ARGS = {
    'model': model
}
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 float, ndim 0, dtype family float, 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
compute_proxy_target_loss Ran
ran on a synthesized input
lingxiaoli94/pol/pol/utils/union_sdf_loss.py
code served (permissive licence) · get_code("af2296dbabc9067f")
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 torch

torch.manual_seed(0)

# Create small tensors for proxy_val and target_sdf_val
# proxy_val: predicted SDF values, shape (2, 4)
proxy_val = torch.randn(2, 4)

# target_sdf_val: ground truth SDF values, shape (2, 4)
target_sdf_val = torch.randn(2, 4)

ARGS = {
    "proxy_val": proxy_val,
    "target_sdf_val": target_sdf_val,
}
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 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
compute_proxy_union_sdf Ran
ran on a synthesized input
lingxiaoli94/pol/pol/utils/union_sdf_loss.py
code served (permissive licence) · get_code("7d19aa8b0e714f75")
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 takes union_sdf_vals as a list/sequence of tensors
# It stacks them along dimension -1, then takes the min along that dimension
# So we need a list of tensors with compatible shapes

# Create a few small SDF value tensors
# Each tensor represents SDF values, typically shape (batch, spatial_dims...)
# Let's make them (batch=2, spatial=4) for simplicity

sdf_tensor_1 = torch.randn(2, 4)
sdf_tensor_2 = torch.randn(2, 4)
sdf_tensor_3 = torch.randn(2, 4)

union_sdf_vals = [sdf_tensor_1, sdf_tensor_2, sdf_tensor_3]

ARGS = {
    "union_sdf_vals": union_sdf_vals
}
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
create_nd_dataset Ran
ran on a synthesized input
lingxiaoli94/pol/pol/datasets/linear_regression.py
code served (permissive licence) · get_code("2dc8a4235958f08c")
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 TensorDataset. 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 numpy as np
import torch
from torch.utils.data import TensorDataset

torch.manual_seed(0)

ARGS = {
    "dim": 4,
    "num_sample": 2,
    "seed": 1234,
    "noise": 0.1,
    "sparsity": 1.0,
    "weight_bound": 1,
    "torch_type": torch.float32,
}
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 TensorDataset — a TensorDataset 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
create_toy1_2d_dataset Ran
ran on a synthesized input
lingxiaoli94/pol/pol/datasets/linear_regression.py
code served (permissive licence) · get_code("b67eecbc6e345636")
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 TensorDataset. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
from torch.utils.data import TensorDataset

torch.manual_seed(0)

ARGS = {
    "torch_type": torch.float32
}
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 TensorDataset — a TensorDataset 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
get_square_padding Ran
ran on a synthesized input
lingxiaoli94/pol/pol/datasets/objdetect.py
code served (permissive licence) · get_code("8217e92ab2b14394")
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 int; 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 PIL import Image

# Create a small test image
# The function calls image.size which returns (width, height)
# Create an image with distinct dimensions: 4 wide, 3 tall
image = Image.new('RGB', (4, 3))

ARGS = {
    'image': image,
    'top_left': True
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, numpy — 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 int — 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
make_analytical_loss_fn Ran
ran on a synthesized input
lingxiaoli94/pol/pol/datasets/analytical.py
code served (permissive licence) · get_code("60f74c28a1526d90")
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 torch.nn as nn

torch.manual_seed(0)

class StandInProb:
    """Stand-in probability object with eval_fn method."""
    def eval_fn(self, X, theta):
        """
        Evaluates the probability/loss function.
        X: input tensor (batch_size, input_dim)
        theta: parameters tensor (param_dim,)
        Returns: scalar loss value
        """
        # Simple analytical loss: sum of squared differences
        return torch.sum((X @ theta[:X.shape[1]]) ** 2)

# Create small tensors with distinct dimensions
X = torch.randn(2, 4)  # batch_size=2, input_dim=4
theta = torch.randn(5)  # param_dim=5

# Create the probability object
prob = StandInProb()

ARGS = {
    "prob": prob
}
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
make_linear_regression_loss_fn Ran
ran on a synthesized input
lingxiaoli94/pol/pol/datasets/linear_regression.py
code served (permissive licence) · get_code("ce2471f80a234495")
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 torch.nn as nn

torch.manual_seed(0)

class Regularizer:
    """Stand-in regularizer with a regularize_fn method."""
    def regularize_fn(self, W, lambdas):
        """
        Compute regularization penalty.
        W: weight tensor
        lambdas: regularization coefficients
        Returns: regularization loss (scalar or compatible shape)
        """
        # Simple L2 regularization: lambda * ||W||^2
        return (lambdas * (W ** 2).sum()).squeeze()

# Create a regularizer instance
regularizer = Regularizer()

ARGS = {
    "regularizer": regularizer
}
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
max_by_axis Ran
ran on a synthesized input
lingxiaoli94/pol/pol/datasets/objdetect.py
code served (permissive licence) · get_code("0406a3c440bc60ac")
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 int; 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 max_by_axis takes a list of sublists and finds the maximum value
# at each index across all sublists. It modifies the first sublist in-place.
# 
# the_list: a list where:
#   - the_list[0] is the initial maxes (will be modified)
#   - the_list[1:] are sublists to compare element-wise

# Create a simple test case with lists of numbers
the_list = [
    [1, 2, 3, 4],      # initial maxes
    [5, 1, 2, 3],      # sublist 1
    [2, 6, 1, 5],      # sublist 2
    [3, 3, 4, 2]       # sublist 3
]

ARGS = {
    "the_list": the_list
}
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 list, ndim 1, dtype family int — 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

Also present in other papers' code

Of 11 samples lifted from this paper's linked repositories, 11 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 2026-11-22.

Repositories linked to this paper

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

Abstract

Finding multiple solutions of non-convex optimization problems is a ubiquitous yet challenging task. Most past algorithms either apply single-solution optimization methods from multiple random initial guesses or search in the vicinity of found solutions using ad hoc heuristics. We present an end-to-end method to learn the proximal operator of a family of training problems so that multiple local minima can be quickly obtained from initial guesses by iterating the learned operator, emulating the proximal-point algorithm that has fast convergence. The learned proximal operator can be further generalized to recover multiple optima for unseen problems at test time, enabling applications such as object detection. The key ingredient in our formulation is a proximal regularization term, which elevates the convexity of our training loss: by applying recent theoretical results, we show that for weakly-convex objectives with Lipschitz gradients, training of the proximal operator converges globally with a practical degree of over-parameterization. We further present an exhaustive benchmark for multi-solution optimization to demonstrate the effectiveness of our method.

For agents

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

get_harvested_code_for_paper("2201.11945")
get_code_for_paper("2201.11945")
have("2201.11945")

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

curl https://syntology.ai/api/ran/2201.11945.json

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

[![Syntology run record](https://syntology.ai/api/ran/2201.11945.svg)](https://syntology.ai/paper/2201.11945)

Connect an agent — have() is free.