SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2204.01723 · 2022

Signal Propagation: A Framework for Learning and Inference In a Forward Pass

arXiv · PDF · Open in the Atlas

Code that ran

6 of 9 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 6 have no contract verdict; 0 failures of our instrument; 3 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/2204.01723.json · badge: /api/ran/2204.01723.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
amassivek/signalpropagation canonical (classifier label)
classifier basis not recorded on the link
6 ran · 3 no recorded run
FunctionStatusWhere it lives
fixed Ran
ran on a synthesized input
amassivek/signalpropagation/sigprop/propagators/functional_other.py
code served (permissive licence) · get_code("93195d66280788f5")
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 function that applies a linear transformation
def simple_func(x):
    return x * 2.0

# Create input components
h0 = torch.randn(2, 4)  # batch=2, dim=4
t0 = torch.randn(2, 3)  # batch=2, different dim=3 to distinguish
context = {"key": "value"}  # arbitrary context object

input_tuple = (h0, t0, context)

ARGS = {
    "input": input_tuple,
    "func": simple_func
}
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
forward Ran
ran on a synthesized input
amassivek/signalpropagation/sigprop/propagators/functional_other.py
code served (permissive licence) · get_code("c2297c3c48ddf90e")
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 function that processes tensors
class SimpleFunc(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(4, 3)
    
    def forward(self, x):
        if x is None:
            return None
        return self.linear(x)

# Instantiate the function
func = SimpleFunc()

# Create input tensors with distinct shapes
h0 = torch.randn(2, 4)  # batch_size=2, dim=4
t0 = torch.randn(2, 4)  # batch_size=2, dim=4
context = {"key": "value"}  # arbitrary context

# Create the input tuple
input_data = (h0, t0, context)

ARGS = {
    "input": input_data,
    "func": func
}
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
identity Ran
ran on a synthesized input
amassivek/signalpropagation/sigprop/propagators/functional_other.py
code served (permissive licence) · get_code("cbea6ee53ba108c1")
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)

# Define a simple function that will be passed as the 'func' parameter
def simple_func(x):
    """A simple function that applies a linear transformation"""
    return x * 2 + 1

# Create the input tuple (h0, t0, context)
h0 = torch.randn(2, 4)  # batch size 2, dimension 4
t0 = torch.tensor([1.0, 2.0])  # time step or similar, batch size 2
context = {"key": "value", "data": torch.randn(2, 3)}  # arbitrary context dict

input_tuple = (h0, t0, context)

ARGS = {
    "input": input_tuple,
    "func": simple_func
}
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
similarity_format Ran
ran on a synthesized input
amassivek/signalpropagation/sigprop/loss/functional/utils.py
code served (permissive licence) · get_code("c64701ee09bbb86c")
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 with distinct dimensions
# The function uses x.dim(), x.flatten(), x.size(), x.mean(), torch.sum()
# Let's create a tensor with shape (2, 4, 3, 5) to test the flattening logic
x = torch.randn(2, 4, 3, 5)

ARGS = {
    "x": x,
    "dim": 1,
    "no_std": 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 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
similarity_matrix_x Ran
ran on a synthesized input
amassivek/signalpropagation/sigprop/loss/functional/utils.py
code served (permissive licence) · get_code("24b2a56dbdd01ceb")
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, 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)

# Create a simple tensor for testing
# Shape: (batch=2, dim=4, extra=3) - will be used with dim=1
x = torch.randn(2, 4, 3)

ARGS = {
    "x": x,
    "dim": 1,
}
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, 2) — shape (2, 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
v14_input_target_max_rand Ran
ran on a synthesized input
amassivek/signalpropagation/sigprop/loss/functional/input_target.py
code served (permissive licence) · get_code("d05a2491389f3dbe")
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 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)

# Analyze the function:
# - sp_learn: not used in the function body
# - h1: tensor, flattened and permuted or flattened to (batch, features)
# - t1: tensor, flattened and permuted or flattened to (batch, features)
# - h0: not used in the function body
# - t0: tensor, flattened or flattened/permuted, used for argmax(1)
# - context: not used in the function body

# The function has two paths based on tensor shapes:
# Path 1 (len(t0.shape) == 2): t0 is 2D, produces yy via t0 @ t0.t()
# Path 2 (len(t0.shape) > 2): t0 is 3D, produces yy via mean of batched matmul

# For t1/h1: if len(t1.shape) > 2, they're 3D; otherwise 2D

# Let's construct tensors for the 2D path (simpler):
# t0: shape (batch=2, features=4) -> yy shape (4, 4) -> argmax(1) gives (4,)
# h1: shape (batch=2, features=4) 
# t1: shape (batch=2, features=4)
# y: shape (2, 4) after h1 @ t1.t()
# cross_entropy(y, target) where target shape is (4,) - this will fail dimension check

# Actually, looking at cross_entropy(y, yy.argmax(1)):
# yy.argmax(1) in 2D case: yy is (4, 4), argmax(1) gives shape (4,)
# y should be (4, num_classes) and target should be (4,)
# So y needs to have shape matching yy.argmax(1)

# Let me reconsider: for 2D path:
# t0: (2, 4) -> yy = (4, 4) -> yy.argmax(1) = (4,) with values in [0,3]
# h1: (2, 4), t1: (2, 4) -> y = (2, 4) 
# cross_entropy expects (N, C) and (N,), so y=(2,4) and target=(4,) won't work

# Let's try 3D path instead:
# t0: (spatial=3, batch=2, features=4) -> flattened to (3, 2, 4) -> (3, 2, 4) @ (3, 4, 2) -> (3, 2, 2) -> mean -> (2, 2)
# yy.argmax(1) = (2,) with values in [0,1]
# h1: (spatial=3, batch=2, features=4) -> (3, 2, 4)
# t1: (spatial=3, batch=2, features=4) -> (3, 4, 2) after permute(0, 2, 1)
# y = (3, 2, 2) -> mean -> (2, 2)
# cross_entropy((2, 2), (2,)) works if target values in [0, 1]

sp_learn = None

# 3D tensors: (spatial_dim, batch, features)
h1 = torch.randn(3, 2, 4)
t1 = torch.randn(3, 2, 4)
h0 = torch.randn(3, 2, 4)
t0 = torch.randn(3, 2, 4)

context = None

ARGS = {
    "sp_learn": sp_learn,
    "h1": h1,
    "t1": t1,
    "h0": h0,
    "t0": t0,
    "context": context,
}
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 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
similarity_matrix_xy Not yet run
unverified (harvested, no recorded run)
amassivek/signalpropagation/sigprop/loss/functional/utils.py
code served (permissive licence) · get_code("ec5c90d5663f47f5")
How it ran

no invocation recorded.

v1_input_label_direct Not yet run
unverified (harvested, no recorded run)
amassivek/signalpropagation/sigprop/loss/functional/input_label.py
code served (permissive licence) · get_code("cf52d04921ad7f50")
How it ran

no invocation recorded.

v9_input_target_max_all Not yet run
unverified (harvested, no recorded run)
amassivek/signalpropagation/sigprop/loss/functional/input_target.py
code served (permissive licence) · get_code("13c1dbbc1efd1b13")
How it ran

no invocation recorded.

Also present in other papers' code

Of 9 samples lifted from this paper's linked repositories, 9 carry a function identity and 0 do not. Of the 9 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 9 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

We propose a new learning framework, signal propagation (sigprop), for propagating a learning signal and updating neural network parameters via a forward pass, as an alternative to backpropagation. In sigprop, there is only the forward path for inference and learning. So, there are no structural or computational constraints necessary for learning to take place, beyond the inference model itself, such as feedback connectivity, weight transport, or a backward pass, which exist under backpropagation based approaches. That is, sigprop enables global supervised learning with only a forward path. This is ideal for parallel training of layers or modules. In biology, this explains how neurons without feedback connections can still receive a global learning signal. In hardware, this provides an approach for global supervised learning without backward connectivity. Sigprop by construction has compatibility with models of learning in the brain and in hardware than backpropagation, including alternative approaches relaxing learning constraints. We also demonstrate that sigprop is more efficient in time and memory than they are. To further explain the behavior of sigprop, we provide evidence that sigprop provides useful learning signals in context to backpropagation. To further support relevance to biological and hardware learning, we use sigprop to train continuous time neural networks with Hebbian updates, and train spiking neural networks with only the voltage or with biologically and hardware compatible surrogate functions.

For agents

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

get_harvested_code_for_paper("2204.01723")
get_code_for_paper("2204.01723")
have("2204.01723")

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

curl https://syntology.ai/api/ran/2204.01723.json

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

[![Syntology run record](https://syntology.ai/api/ran/2204.01723.svg)](https://syntology.ai/paper/2204.01723)

Connect an agent — have() is free.