SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2212.01548 · 2022

FedRolex: Model-Heterogeneous Federated Learning with Rolling Sub-Model Extraction

arXiv · PDF · Open in the Atlas

Code that ran

7 of 15 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 3 have no contract verdict; 3 failures of our instrument; 8 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/2212.01548.json · badge: /api/ran/2212.01548.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
aiot-mlsys-lab/fedrolex canonical (classifier label)
classifier basis not recorded on the link
2 ran · 1 no recorded run
k1l1/SLT reimplementation (classifier label)
classifier basis: classifier_2025
5 ran · 6 no recorded run
msu-mlsys-lab/fedrolex alias (classifier label)
classifier basis: classifier_2025
0 ran · 1 no recorded run
FunctionStatusWhere it lives
FedAvgDevice Ran
ran on a synthesized input · constructed an object; no result was computed
k1l1/SLT/algorithms/fedrolex.py
code served (permissive licence) · get_code("c0cad3dd46b90b51")
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, FedAvgDevice: 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 torch.optim as optim
from torch.utils.data import TensorDataset

torch.manual_seed(0)

# Create a minimal neural network model
class SimpleModel(nn.Module):
    def __init__(self, input_dim=4, hidden_dim=8, num_classes=3):
        super().__init__()
        self.fc1 = nn.Linear(input_dim, hidden_dim)
        self.fc2 = nn.Linear(hidden_dim, num_classes)
    
    def forward(self, x):
        x = torch.relu(self.fc1(x))
        x = self.fc2(x)
        return x

# Create small dummy datasets
def create_dataset(n_samples=10, input_dim=4, num_classes=3):
    X = torch.randn(n_samples, input_dim)
    y = torch.randint(0, num_classes, (n_samples,))
    return TensorDataset(X, y)

train_dataset = create_dataset(n_samples=10, input_dim=4, num_classes=3)
test_dataset = create_dataset(n_samples=8, input_dim=4, num_classes=3)

# Model class and kwargs
model_class = SimpleModel
model_kwargs = {'input_dim': 4, 'hidden_dim': 8, 'num_classes': 3}

# Optimizer and loss function
optimizer_class = optim.SGD
optimizer_args = {'momentum': 0.9}

loss_function = nn.CrossEntropyLoss

# Torch device
torch_device = torch.device('cpu')

ARGS = {
    'device_id': 0,
    'storage_path': '/tmp/fed_avg_test',
}
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 FedAvgDevice — 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 FedAvgDevice 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
FedAvgEvaluation Ran
ran on a synthesized input · constructed an object; no result was computed
k1l1/SLT/algorithms/fedrolex.py
code served (permissive licence) · get_code("ae7e474951511463")
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, FedAvgEvaluation: 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 torch.utils.data import TensorDataset, DataLoader

torch.manual_seed(0)

# Create a simple model
class SimpleModel(nn.Module):
    def __init__(self, input_dim=4, num_classes=3):
        super(SimpleModel, self).__init__()
        self.fc1 = nn.Linear(input_dim, 8)
        self.fc2 = nn.Linear(8, num_classes)
    
    def forward(self, x):
        x = torch.relu(self.fc1(x))
        x = self.fc2(x)
        return x

# Create dummy train dataset
train_inputs = torch.randn(10, 4)
train_labels = torch.randint(0, 3, (10,))
train_dataset = TensorDataset(train_inputs, train_labels)
train_loader = DataLoader(train_dataset, batch_size=2)

# Create dummy test dataset with train flag
class TestDataset(TensorDataset):
    def __init__(self, *tensors):
        super().__init__(*tensors)
        self.train = False

test_inputs = torch.randn(8, 4)
test_labels = torch.randint(0, 3, (8,))
test_dataset = TestDataset(test_inputs, test_labels)
test_loader = DataLoader(test_dataset, batch_size=2)

# Model class and kwargs
model_class = SimpleModel
model_kwargs = {"input_dim": 4, "num_classes": 3}

# Optimizer
optimizer_class = torch.optim.SGD
optimizer_args = {"momentum": 0.9}

# Loss function
loss_function = nn.CrossEntropyLoss

# Torch device
torch_device = torch.device('cpu')

ARGS = {
    'device_id': 0,
    'storage_path': '/tmp/test_device'
}

# Additional setup after instantiation
device = None

def setup_fedavg_evaluation():
    from inspect import signature
    
    eval_device = type('FedAvgEvaluation', (), {
        '__init__': lambda self, device_id, storage_path: (
            setattr(self, '_device_id', device_id),
            setattr(self, '_storage_path', storage_path),
            setattr(self, '_model', None),
            setattr(self, '_model_kwargs', None),
            setattr(self, '_model_class', None),
            setattr(self, '_test_data', None),
            setattr(self, '_train_data', None),
            setattr(self, '_batch_size_test', 1024),
            setattr(self, '_batch_size_train', 32),
            setattr(self, '_optimizer', None),
            setattr(self, '_optimizer_args', None),
            setattr(self, '_loss_F', None),
            setattr(self, 'lr', 0.01),
            setattr(self, 'training_loss', None),
            setattr(self, '_torch_device', None),
            setattr(self, '_accuracy_test', None),
        )
    })()
    
    eval_device.set_seed(0)
    eval_device.set_model(model_class, model_kwargs)
    eval_device.init_model()
    eval_device.set_test_data(test_loader)
    eval_device.set_train_data(train_loader)
    eval_device.set_torch_device(torch_device)
    eval_device.set_optimizer(optimizer_class, optimizer_args)
    eval_device.set_loss_function(loss_function)
    
    return eval_device

ARGS = {
    'device_id': 0,
    'storage_path': '/tmp/test_device'
}
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 FedAvgEvaluation — 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 FedAvgEvaluation 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
dict_hash Ran
ran; our contract draft was wrong, not the code
k1l1/SLT/run_fl.py
code served (permissive licence) · get_code("35d5b46e4f624d34")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Sonnet 4.5, prompt python_fixture_v1) and returned str. 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 hashlib
import json

# The function takes a dictionary and computes its MD5 hash
# We need to provide a valid dictionary that can be JSON-serialized

dictionary = {
    "key1": "value1",
    "key2": 42,
    "key3": [1, 2, 3],
    "key4": {"nested": "dict"},
    "key5": True,
    "key6": None
}

ARGS = {
    "dictionary": dictionary
}
Setup block drafted by
a language model, Claude Sonnet 4.5, prompt python_fixture_v1
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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
extract_submodel_resnet_structure Ran
ran; our contract draft was wrong, not the code
k1l1/SLT/utils/fedrolex_submodel.py
code served (permissive licence) · get_code("558b018bc1c0f19a")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Sonnet 4.5, prompt python_fixture_v1) and returned tuple, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np
import copy

torch.manual_seed(0)

# Create a minimal ResNet-like state dict structure
# The function expects a state dict (OrderedDict or dict) with specific layer naming patterns

global_model = {}

# First conv layer: input channels = 3 (RGB), output channels = 8
global_model['conv1.weight'] = torch.randn(8, 3, 3, 3)
global_model['bn1.weight'] = torch.randn(8)
global_model['bn1.bias'] = torch.randn(8)
global_model['bn1.running_mean'] = torch.randn(8)
global_model['bn1.running_var'] = torch.randn(8)
global_model['bn1.num_batches_tracked'] = torch.tensor(100)

# Layer1 block1
global_model['layer1.0.conv1.weight'] = torch.randn(8, 8, 3, 3)
global_model['layer1.0.bn1.weight'] = torch.randn(8)
global_model['layer1.0.bn1.bias'] = torch.randn(8)
global_model['layer1.0.bn1.running_mean'] = torch.randn(8)
global_model['layer1.0.bn1.running_var'] = torch.randn(8)
global_model['layer1.0.bn1.num_batches_tracked'] = torch.tensor(100)

global_model['layer1.0.conv2.weight'] = torch.randn(8, 8, 3, 3)
global_model['layer1.0.bn2.weight'] = torch.randn(8)
global_model['layer1.0.bn2.bias'] = torch.randn(8)
global_model['layer1.0.bn2.running_mean'] = torch.randn(8)
global_model['layer1.0.bn2.running_var'] = torch.randn(8)
global_model['layer1.0.bn2.num_batches_tracked'] = torch.tensor(100)

# Shortcut connection
global_model['layer1.0.shortcut.conv.weight'] = torch.randn(8, 8, 1, 1)
global_model['layer1.0.shortcut.bn.weight'] = torch.randn(8)
global_model['layer1.0.shortcut.bn.bias'] = torch.randn(8)
global_model['layer1.0.shortcut.bn.running_mean'] = torch.randn(8)
global_model['layer1.0.shortcut.bn.running_var'] = torch.randn(8)
global_model['layer1.0.shortcut.bn.num_batches_tracked'] = torch.tensor(100)

# Layer2 block1
global_model['layer2.0.conv1.weight'] = torch.randn(16, 8, 3, 3)
global_model['layer2.0.bn1.weight'] = torch.randn(16)
global_model['layer2.0.bn1.bias'] = torch.randn(16)
global_model['layer2.0.bn1.running_mean'] = torch.randn(16)
global_model['layer2.0.bn1.running_var'] = torch.randn(16)
global_model['layer2.0.bn1.num_batches_tracked'] = torch.tensor(100)

global_model['layer2.0.conv2.weight'] = torch.randn(16, 16, 3, 3)
global_model['layer2.0.bn2.weight'] = torch.randn(16)
global_model['layer2.0.bn2.bias'] = torch.randn(16)
global_model['layer2.0.bn2.running_mean'] = torch.randn(16)
global_model['layer2.0.bn2.running_var'] = torch.randn(16)
global_model['layer2.0.bn2.num_batches_tracked'] = torch.tensor(100)

# Linear layer (fully connected)
# Assuming feature map size after conv layers is 2x2, and we have 16 channels
# So flattened size = 16 * 2 * 2 = 64
global_model['linear.weight'] = torch.randn(10, 64)
global_model['linear.bias'] = torch.randn(10)

ratio = 0.5
round_n = 2

ARGS = {
    'ratio': ratio,
    'global_model': global_model,
    'round_n': round_n
}
Setup block drafted by
a language model, Claude Sonnet 4.5, prompt python_fixture_v1
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
input_collate Ran
ran; our contract draft was wrong, not the code
aiot-mlsys-lab/fedrolex/data.py
code served (permissive licence) · get_code("6fd4f7ea4a79121e")
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 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
import numpy as np
from torch.utils.data.dataloader import default_collate

torch.manual_seed(0)

# Test case 1: batch of dictionaries
batch_dicts = [
    {"input_ids": torch.tensor([1, 2, 3, 4]), "attention_mask": torch.tensor([1, 1, 1, 0])},
    {"input_ids": torch.tensor([5, 6, 7, 8]), "attention_mask": torch.tensor([1, 1, 1, 1])}
]

ARGS = {
    "batch": batch_dicts
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, 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
make_node Ran
ran on a synthesized input
aiot-mlsys-lab/fedrolex/datasets/imagenet.py
code served (permissive licence) · get_code("b81843a16c748c02")
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 numpy as np
import torch

# Create node structure that .item() returns a tuple with at least 3 elements
# The function expects node.item() to return (id, wnid, classes, ..., children_info)
# where children_info is at index 5

# Create a simple numpy array that when .item() is called returns a tuple
node_data = np.array([(1, 'n0001', 'entity', 0, 0, np.array([2, 3]), 0)], dtype=object)
node = node_data[0]

# Create branch as a list of numpy arrays where each can be indexed
# Each branch element needs .item() that returns a tuple with children info at index 5
branch_elem_1 = np.array([(2, 'n0002', 'entity', 0, 0, 1, 0)], dtype=object)[0]
branch_elem_2 = np.array([(3, 'n0003', 'entity', 0, 0, 1, 0)], dtype=object)[0]
branch_elem_3 = np.array([(4, 'n0004', 'entity', 0, 0, np.array([]), 0)], dtype=object)[0]

# Wrap each in an object that has .item() method
class NodeWrapper:
    def __init__(self, data):
        self.data = data
    
    def item(self):
        return self.data

branch = [
    NodeWrapper((2, 'n0002', 'entity', 0, 0, 1, 0)),
    NodeWrapper((3, 'n0003', 'entity', 0, 0, 1, 0)),
    NodeWrapper((4, 'n0004', 'entity', 0, 0, np.array([]), 0))
]

node_wrapped = NodeWrapper((1, 'n0001', 'entity', 0, 0, np.array([2, 3]), 0))

ARGS = {
    'node': node_wrapped,
    'branch': branch
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — 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
submodel_indices Ran
ran, honoured the contract we drafted
k1l1/SLT/utils/fedrolex_submodel.py
code served (permissive licence) · get_code("11c9c63eba295e97")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Sonnet 4.5, prompt python_fixture_v1) and returned Tensor, ndim 1, dtype family int, shape (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 numpy as np

torch.manual_seed(0)

round_n = 3
ratio = 0.6
shape = 10

ARGS = {
    'round_n': round_n,
    'ratio': ratio,
    'shape': shape
}
Setup block drafted by
a language model, Claude Sonnet 4.5, prompt python_fixture_v1
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 int, shape (6,) — shape (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
FedAvgServer Not yet run
unverified (harvested, no recorded run)
k1l1/SLT/algorithms/fedrolex.py
code served (permissive licence) · get_code("b34975803f9db84d")
How it ran

no invocation recorded.

FedRolexServer Not yet run
unverified (harvested, no recorded run)
k1l1/SLT/algorithms/fedrolex.py
code served (permissive licence) · get_code("39db5b4010e71eb9")
How it ran

no invocation recorded.

FederatedDropoutServer Not yet run
unverified (harvested, no recorded run)
k1l1/SLT/algorithms/fedrolex.py
code served (permissive licence) · get_code("b7c57b09b48aa5d6")
How it ran

no invocation recorded.

ResnetServerRoll Not yet run
unverified (harvested, no recorded run)
msu-mlsys-lab/fedrolex/resnet_server.py
code served (permissive licence) · get_code("c03c7b2e83fd0bcd")
How it ran

no invocation recorded.

SubsetDevice Not yet run
unverified (harvested, no recorded run)
k1l1/SLT/algorithms/fedrolex.py
code served (permissive licence) · get_code("246b2e9dbd5d9a20")
How it ran

no invocation recorded.

extract_submodel_densenet_structure Not yet run
unverified (harvested, no recorded run)
k1l1/SLT/utils/fedrolex_submodel.py
code served (permissive licence) · get_code("cd24f922c541136d")
How it ran

no invocation recorded.

fetch_dataset Not yet run
unverified (harvested, no recorded run)
aiot-mlsys-lab/fedrolex/data.py
code served (permissive licence) · get_code("1231ff80106c5e8e")
How it ran

no invocation recorded.

prepare Not yet run
unverified (harvested, no recorded run)
k1l1/SLT/run_fl.py
code served (permissive licence) · get_code("22f89a73c84ff63c")
How it ran

no invocation recorded.

Also present in other papers' code

Of 15 samples lifted from this paper's linked repositories, 15 carry a function identity and 0 do not. Of the 15 distinct functions with an identity, 8 are also present in other papers whose code we hold (7 with at least one paper dated earlier, 0 the same month, 8 later) and 7 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
FedAvgDevice
brick of this paper's assembly: helper
3 other papers
2 dated earlier, 1 later · 1 only as an assembly brick
2010.01264 (2020-10)
HeteroFL: Computation and Communication Efficient Federated Learning for Heterogeneous Clients
2102.13451 (2021-02)
FjORD: Fair and Accurate Federated Learning under heterogeneous targets with Ordered Dropout
2305.17005 (2023-05)
Aggregating Capacity in FL through Successive Layer Training for Computationally-Constrained Devices
SubsetDevice
brick of this paper's assembly: helper
3 other papers
2 dated earlier, 1 later · 1 only as an assembly brick
2010.01264 (2020-10)
HeteroFL: Computation and Communication Efficient Federated Learning for Heterogeneous Clients
2102.13451 (2021-02)
FjORD: Fair and Accurate Federated Learning under heterogeneous targets with Ordered Dropout
2305.17005 (2023-05)
Aggregating Capacity in FL through Successive Layer Training for Computationally-Constrained Devices
input_collate 3 other papers
2 dated earlier, 1 later
2010.01264 (2020-10)
HeteroFL: Computation and Communication Efficient Federated Learning for Heterogeneous Clients
2106.01425 (2021-06)
GAL: Gradient Assisted Learning for Decentralized Multi-Organization Collaborations
2405.19789 (2024-05)
Estimating before Debiasing: A Bayesian Approach to Detaching Prior Bias in Federated Semi-Supervised Learning
FedAvgEvaluation
brick of this paper's assembly: helper
2 other papers
1 dated earlier, 1 later
2010.01264 (2020-10)
HeteroFL: Computation and Communication Efficient Federated Learning for Heterogeneous Clients
2305.17005 (2023-05)
Aggregating Capacity in FL through Successive Layer Training for Computationally-Constrained Devices
FedAvgServer
brick of this paper's assembly: helper
2 other papers
1 dated earlier, 1 later
2010.01264 (2020-10)
HeteroFL: Computation and Communication Efficient Federated Learning for Heterogeneous Clients
2305.17005 (2023-05)
Aggregating Capacity in FL through Successive Layer Training for Computationally-Constrained Devices
FederatedDropoutServer
brick of this paper's assembly: helper
2 other papers
1 dated earlier, 1 later
2010.01264 (2020-10)
HeteroFL: Computation and Communication Efficient Federated Learning for Heterogeneous Clients
2305.17005 (2023-05)
Aggregating Capacity in FL through Successive Layer Training for Computationally-Constrained Devices
dict_hash 2 other papers
1 dated earlier, 1 later
2203.05468 (2022-03)
CoCoFL: Communication- and Computation-Aware Federated Learning via Partial NN Freezing and Quantization
2305.17005 (2023-05)
Aggregating Capacity in FL through Successive Layer Training for Computationally-Constrained Devices
prepare 1 other paper
1 later
2305.17005 (2023-05)
Aggregating Capacity in FL through Successive Layer Training for Computationally-Constrained Devices

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

Most cross-device federated learning (FL) studies focus on the model-homogeneous setting where the global server model and local client models are identical. However, such constraint not only excludes low-end clients who would otherwise make unique contributions to model training but also restrains clients from training large models due to on-device resource bottlenecks. In this work, we propose FedRolex, a partial training (PT)-based approach that enables model-heterogeneous FL and can train a global server model larger than the largest client model. At its core, FedRolex employs a rolling sub-model extraction scheme that allows different parts of the global server model to be evenly trained, which mitigates the client drift induced by the inconsistency between individual client models and server model architectures. We show that FedRolex outperforms state-of-the-art PT-based model-heterogeneous FL methods (e.g. Federated Dropout) and reduces the gap between model-heterogeneous and model-homogeneous FL, especially under the large-model large-dataset regime. In addition, we provide theoretical statistical analysis on its advantage over Federated Dropout and evaluate FedRolex on an emulated real-world device distribution to show that FedRolex can enhance the inclusiveness of FL and boost the performance of low-end devices that would otherwise not benefit from FL. Our code is available at: https://github.com/AIoT-MLSys-Lab/FedRolex

For agents

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

get_harvested_code_for_paper("2212.01548")
get_code_for_paper("2212.01548")
have("2212.01548")

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

curl https://syntology.ai/api/ran/2212.01548.json

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

[![Syntology run record](https://syntology.ai/api/ran/2212.01548.svg)](https://syntology.ai/paper/2212.01548)

Connect an agent — have() is free.