SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2105.04100 · ICML · 2021

Z-GCNETs: Time Zigzags at Graph Convolutional Networks for Time Series Forecasting

Yuzhou Chen, Ignacio Segovia-Dominguez, Yulia Gel

arXiv · PDF · Open in the Atlas

Code that ran

3 of 4 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (3 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 3 have no contract verdict; 0 failures of our instrument; 1 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/2105.04100.json · badge: /api/ran/2105.04100.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
Z-GCNETs/Z-GCNETs canonical (classifier label)
classifier basis: classifier_2025
3 ran · 1 no recorded run
FunctionStatusWhere it lives
CNN Ran
ran on a synthesized input · constructed an object; no result was computed
Z-GCNETs/Z-GCNETs/model/ZGCNETs.py
code served (permissive licence) · get_code("ace626efd2e7365c")
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, CNN: the run constructed the object and did not call it, so no result was computed. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# The CNN class expects:
# - dim_out: an integer for output dimension

ARGS = {
    "dim_out": 16
}
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 CNN — 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 CNN 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
NNLSDCRNNCNN Ran
ran on a synthesized input · constructed an object; no result was computed
Z-GCNETs/Z-GCNETs/model/ZGCNETs.py
code served (permissive licence) · get_code("0250d58e6f6bdde2")
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, NNLSDCRNNCNN: the run constructed the object and did not call it, so no result was computed. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# Define minimal stand-in classes and tensors
node_num = 5
dim_in = 3
dim_out = 8
link_len = 2
embed_dim = 4
num_layers = 2
window_len = 12

# Create node embeddings (node_num x embed_dim)
node_embeddings = torch.randn(node_num, embed_dim)

# Create zigzag_PI tensor - used as input to CNN which expects (batch, channels, height, width)
# CNN expects 1 channel input, so we create (batch_size, 1, height, width)
batch_size = 2
zigzag_PI = torch.randn(batch_size, 1, 28, 28)

# Create initialization states for each layer
# Each state should be (batch_size, node_num, dim_out)
init_state = [torch.randn(batch_size, node_num, dim_out) for _ in range(num_layers)]

# Create input tensor x with shape (batch_size, seq_length, node_num, input_dim)
seq_length = 6
x = torch.randn(batch_size, seq_length, node_num, dim_in)

ARGS = {
    'node_num': node_num,
    'dim_in': dim_in,
    'dim_out': dim_out,
    'link_len': link_len,
    'embed_dim': embed_dim,
    'num_layers': num_layers,
    'window_len': window_len
}
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 NNLSDCRNNCNN — 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 NNLSDCRNNCNN 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
TFLSGCNCNN Ran
ran on a synthesized input · constructed an object; no result was computed
Z-GCNETs/Z-GCNETs/model/ZGCNETs.py
code served (permissive licence) · get_code("07e683e5ad493643")
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, TFLSGCNCNN: 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.nn.functional as F

torch.manual_seed(0)

class CNN(nn.Module):

    def __init__(self, dim_out):
        super(CNN, self).__init__()
        self.dim_out = dim_out
        self.features = nn.Sequential(nn.Conv2d(1, 8, kernel_size=3, stride=2), nn.ReLU(inplace=True), nn.MaxPool2d(kernel_size=3, stride=2), nn.Conv2d(8, dim_out, kernel_size=3, stride=2), nn.ReLU(inplace=True), nn.MaxPool2d(kernel_size=3, stride=2))
        self.maxpool = nn.MaxPool2d(5, 5)

    def forward(self, zigzag_window_PI):
        feature = self.features(zigzag_window_PI)
        feature = self.maxpool(feature)
        feature = feature.view(-1, self.dim_out)
        return feature

class TFLSGCNCNN(nn.Module):

    def __init__(self, dim_in, dim_out, window_len, link_len, embed_dim):
        super(TFLSGCNCNN, self).__init__()
        self.link_len = link_len
        self.weights_pool = nn.Parameter(torch.FloatTensor(embed_dim, link_len, dim_in, int(dim_out / 2)))
        if (dim_in - 1) % 16 == 0:
            self.weights_window = nn.Parameter(torch.FloatTensor(embed_dim, 1, int(dim_out / 2)))
        else:
            self.weights_window = nn.Parameter(torch.FloatTensor(embed_dim, int(dim_in / 2), int(dim_out / 2)))
        self.bias_pool = nn.Parameter(torch.FloatTensor(embed_dim, dim_out))
        self.T = nn.Parameter(torch.FloatTensor(window_len))
        self.cnn = CNN(int(dim_out / 2))

    def forward(self, x, x_window, node_embeddings, zigzag_PI):
        node_num = node_embeddings.shape[0]
        supports = F.softmax(F.relu(torch.mm(node_embeddings, node_embeddings.transpose(0, 1))), dim=1)
        support_set = [torch.eye(node_num).to(supports.device), supports]
        for k in range(2, self.link_len):
            support_set.append(torch.mm(supports, support_set[k - 1]))
        supports = torch.stack(support_set, dim=0)
        weights = torch.einsum('nd,dkio->nkio', node_embeddings, self.weights_pool)
        bias = torch.matmul(node_embeddings, self.bias_pool)
        x_g = torch.einsum('knm,bmc->bknc', supports, x)
        x_g = x_g.permute(0, 2, 1, 3)
        x_gconv = torch.einsum('bnki,nkio->bno', x_g, weights)
        weights_window = torch.einsum('nd,dio->nio', node_embeddings, self.weights_window)
        x_w = torch.einsum('btni,nio->btno', x_window, weights_window)
        x_w = x_w.permute(0, 2, 3, 1)
        x_wconv = torch.matmul(x_w, self.T)
        topo_cnn = self.cnn(zigzag_PI)
        x_tgconv = torch.einsum('bno,bo->bno', x_gconv, topo_cnn)
        x_twconv = torch.einsum('bno,bo->bno', x_wconv, topo_cnn)
        x_gwconv = torch.cat([x_tgconv, x_twconv], dim=-1) + bias
        return x_gwconv

dim_in = 8
dim_out = 4
window_len = 5
link_len = 3
embed_dim = 6
batch_size = 2
node_num = 6

x = torch.randn(batch_size, node_num, dim_in)
x_window = torch.randn(batch_size, window_len, node_num, dim_in)
node_embeddings = torch.randn(node_num, embed_dim)
zigzag_PI = torch.randn(batch_size, 1, 32, 32)

ARGS = {
    'dim_in': dim_in,
    'dim_out': dim_out,
    'window_len': window_len,
    'link_len': link_len,
    'embed_dim': embed_dim
}
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 TFLSGCNCNN — 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 TFLSGCNCNN 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
NLSGCRNCNNCell Not yet run
unverified (harvested, no recorded run)
Z-GCNETs/Z-GCNETs/model/ZGCNETs.py
code served (permissive licence) · get_code("79e62d9d6ed44ac9")
How it ran

no invocation recorded.

Also present in other papers' code

Of 4 samples lifted from this paper's linked repositories, 4 carry a function identity and 0 do not. Of the 4 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 4 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

There recently has been a surge of interest in developing a new class of deep learning (DL) architectures that integrate an explicit time dimension as a fundamental building block of learning and representation mechanisms. In turn, many recent results show that topological descriptors of the observed data, encoding information on the shape of the dataset in a topological space at different scales, that is, persistent homology of the data, may contain important complementary information, improving both performance and robustness of DL. As convergence of these two emerging ideas, we propose to enhance DL architectures with the most salient time-conditioned topological information of the data and introduce the concept of zigzag persistence into time-aware graph convolutional networks (GCNs). Zigzag persistence provides a systematic and mathematically rigorous framework to track the most important topological features of the observed data that tend to manifest themselves over time. To integrate the extracted time-conditioned topological descriptors into DL, we develop a new topological summary, zigzag persistence image, and derive its theoretical stability guarantees. We validate the new GCNs with a time-aware zigzag topological layer (Z-GCNETs), in application to traffic forecasting and Ethereum blockchain price prediction. Our results indicate that Z-GCNET outperforms 13 state-of-the-art methods on 4 time series datasets.

For agents

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

get_harvested_code_for_paper("2105.04100")
get_code_for_paper("2105.04100")
have("2105.04100")

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

curl https://syntology.ai/api/ran/2105.04100.json

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

[![Syntology run record](https://syntology.ai/api/ran/2105.04100.svg)](https://syntology.ai/paper/2105.04100)

Connect an agent — have() is free.