Skip to content

COMPARISON.md

Latest commit

 

History

History
180 lines (147 loc) · 17.3 KB

COMPARISON.md

File metadata and controls

180 lines (147 loc) · 17.3 KB

pycleora vs Competitors — Feature Comparison

Libraries Compared

Library Language Backend Focus Status
pycleora 2.3 Rust + Python CPU (Rust SpMM) Full-stack graph embedding SDK Active
PyG (PyTorch Geometric) Python GPU (PyTorch) GNN training framework Active
KarateClub Python CPU (numpy/scipy) Traditional graph embedding Active
Cleora (BaseModelAI) Rust CPU Sparse embedding only Minimal
Node2Vec (Eliorc) Python CPU Random walk embedding Active
DGL Python GPU (PyTorch/MXNet) GNN training framework Active
StellarGraph Python GPU (TensorFlow/Keras) GNN & graph ML library Archived
GEM Python CPU (numpy/scipy) Graph embedding methods Inactive
GraphVite C++/Python GPU (CUDA) High-speed graph embedding Inactive
DeepWalk Python CPU (gensim) Random walk + Skip-gram Inactive
LINE C++/Python CPU 1st/2nd order proximity Inactive
SDNE Python GPU (TF/Keras) Autoencoder-based embedding Inactive
graspologic Python CPU (numpy/scipy) Spectral graph statistics Active
GraphSAGE Python GPU (TF) Inductive node embedding Inactive
Struc2Vec Python CPU Structural identity embedding Inactive
VERSE C++/Python CPU Versatile graph embedding Inactive
NetSMF C++/Python CPU Sparse matrix factorization Inactive

Sources: StellarGraph repo (archived 2022), GEM repo (last commit 2020), GraphVite repo (last commit 2020), Node2Vec/eliorc, DeepWalk (last commit 2019), LINE (last commit 2017), SDNE (last commit 2018), graspologic (Microsoft, active), GraphSAGE (reference impl, last commit 2019), Struc2Vec (last commit 2018), VERSE (last commit 2019), NetSMF (last commit 2019). Install sizes estimated from pip download + dependencies. Timing estimates are order-of-magnitude based on published benchmarks and documentation.

Feature Matrix

Embedding Algorithms

Feature pycleora PyG KarateClub Cleora (Base) Node2Vec DGL StellarGraph GEM GraphVite DeepWalk LINE SDNE graspologic GraphSAGE Struc2Vec VERSE NetSMF
Cleora (Markov propagation) Yes No No Yes No No No No No No No No No No No No No
ProNE Yes No Yes No No No No No No No No No No No No No No
RandNE Yes No Yes No No No No No No No No No No No No No No
HOPE Yes No Yes No No No No Yes No No No No No No No No No
NetMF Yes No Yes No No No No No No No No No No No No No No
GraRep Yes No Yes No No No No No No No No No No No No No No
Node2Vec/DeepWalk Yes Yes Yes No Yes No Yes Yes Yes Yes No No No No No No No
LINE No No No No No No No No Yes No Yes No No No No No No
SDNE No No No No No No No Yes No No No Yes No No No No No
Laplacian Eigenmaps No No Yes No No No No Yes No No No No Yes No No No No
GCN/GAT/GraphSAGE No Yes No No No Yes Yes No No No No No No Yes No No No
Attention-weighted embed Yes Partial No No No Partial Partial No No No No No No No No No No
Multiscale embedding Yes No No No No No No No No No No No No No No No No
Supervised refinement Yes Yes No No No Yes Yes No No No No No No Yes No No No
Struc2Vec No No No No No No No No No No No No No No Yes No No
VERSE No No No No No No No No No No No No No No No Yes No
NetSMF No No No No No No No No No No No No No No No No Yes
ASE/LSE (spectral) No No No No No No No No No No No No Yes No No No No
Total algorithms 11 50+ 30+ 1 1 50+ ~15 ~8 ~5 1 1 1 ~5 1 1 1 1

Graph Types

Feature pycleora PyG KarateClub Cleora (Base) Node2Vec DGL StellarGraph GEM GraphVite DeepWalk LINE SDNE graspologic GraphSAGE Struc2Vec VERSE NetSMF
Undirected graphs Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Directed graphs Yes Yes Partial No No Yes Yes Yes Yes No Yes Yes Yes No No Yes No
Weighted edges Yes Yes Partial No Yes Yes Yes Yes Yes No Yes Yes Yes No No Yes Yes
Hypergraph/bipartite Yes Yes No Yes No Yes Yes No No No No No No Yes No No No
Heterogeneous graphs Yes Yes No No No Yes Yes No No No No No No Yes No No No
Node features Yes Yes No No No Yes Yes No No No No No No Yes No No No

Dynamic/Streaming

Feature pycleora PyG KarateClub Cleora (Base) Node2Vec DGL StellarGraph GEM GraphVite DeepWalk LINE SDNE graspologic GraphSAGE Struc2Vec VERSE NetSMF
Incremental update (add edges) Yes No No No No No No No No No No No No No No No No
Edge removal Yes No No No No No No No No No No No No No No No No
Inductive embedding Yes Yes No No No Yes Yes No No No No No No Yes No No No
Streaming (batch-by-batch) Yes No No No No No No No No No No No No No No No No

Classification

Feature pycleora PyG KarateClub Cleora (Base) Node2Vec DGL StellarGraph GEM GraphVite DeepWalk LINE SDNE graspologic GraphSAGE Struc2Vec VERSE NetSMF
Label Propagation Yes Yes No No No No No No No No No No No No No No No
MLP classifier (built-in) Yes No No No No No No No No No No No No No No No No
GNN classifiers No Yes No No No Yes Yes No No No No No No Yes No No No

Community Detection

Feature pycleora PyG KarateClub Cleora (Base) Node2Vec DGL StellarGraph GEM GraphVite DeepWalk LINE SDNE graspologic GraphSAGE Struc2Vec VERSE NetSMF
K-Means clustering Yes No No No No No No No No No No No No No No No No
Spectral clustering Yes No No No No No No No No No No No No No No No No
Louvain algorithm Yes No No No No No No No No No No No No No No No No
Modularity score Yes No No No No No No No No No No No No No No No No

Evaluation Metrics

Feature pycleora PyG KarateClub Cleora (Base) Node2Vec DGL StellarGraph GEM GraphVite DeepWalk LINE SDNE graspologic GraphSAGE Struc2Vec VERSE NetSMF
Link prediction (AUC/MRR/Hits@K) Yes Partial No No No No Partial No Partial No No No No No No No No
Node classification (Acc/F1) Yes Partial Yes No No No Partial No Partial No No No No No No No No
Clustering (NMI/Purity) Yes No No No No No No No No No No No No No No No No

Built-in Datasets

Feature pycleora PyG KarateClub Cleora (Base) Node2Vec DGL StellarGraph GEM GraphVite DeepWalk LINE SDNE graspologic GraphSAGE Struc2Vec VERSE NetSMF
Small graphs (Karate, etc.) 4 4+ 4+ 0 0 4+ 3+ 2 0 0 0 0 3+ 0 0 0 0
Citation networks (Cora, etc.) 4 4+ 0 0 0 4+ 3+ 1 3+ 0 0 0 0 0 0 0 0
Amazon co-purchase 2 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
Large-scale (PPI, Reddit, DBLP) 2 5+ 0 0 0 5+ 4+ 0 2+ 0 0 0 2+ 0 0 0 0
Total datasets 14 70+ ~5 0 0 40+ ~10 ~3 ~5 0 0 0 ~5 0 0 0 0

I/O & Interoperability

Feature pycleora PyG KarateClub Cleora (Base) Node2Vec DGL StellarGraph GEM GraphVite DeepWalk LINE SDNE graspologic GraphSAGE Struc2Vec VERSE NetSMF
NetworkX import/export Yes Yes Yes No Yes Yes Yes Yes No Yes No No Yes No Yes No No
PyG Data export Yes N/A No No No Yes No No No No No No No No No No No
DGL Graph export Yes Yes No No No N/A No No No No No No No No No No No
Save/load NPZ Yes No No No No No No No No No No No No No No No No
Save/load CSV/TSV Yes No No Yes No No No No No No No No No No No No No
Save/load Parquet Yes No No No No No No No No No No No No No No No No
Pickle serialization Yes Yes Yes No Yes Yes Yes Yes No No No No Yes No No No No

Visualization

Feature pycleora PyG KarateClub Cleora (Base) Node2Vec DGL StellarGraph GEM GraphVite DeepWalk LINE SDNE graspologic GraphSAGE Struc2Vec VERSE NetSMF
t-SNE (built-in) Yes No No No No No No No Partial No No No No No No No No
PCA (built-in) Yes No No No No No No No No No No No No No No No No
UMAP support Yes No No No No No No No No No No No No No No No No
Plot with labels/colors Yes No No No No No No No No No No No Yes No No No No

Performance & Setup

Feature pycleora PyG KarateClub Cleora (Base) Node2Vec DGL StellarGraph GEM GraphVite DeepWalk LINE SDNE graspologic GraphSAGE Struc2Vec VERSE NetSMF
No GPU required Yes No Yes Yes Yes No Optional Yes No Yes Yes Optional Yes Optional Yes Yes Yes
Optional GPU support Yes Yes No No No Yes Yes No Yes No No Yes No Yes No No No
Multi-GPU support Planned Yes No No No Yes Limited No Yes No No No No No No No No
Rust-native performance Yes C++/CUDA No Yes No C++/CUDA No No C++/CUDA No C++ No No No No C++ C++
pip install (no build) No Yes Yes No Yes Yes Yes No No Yes No No Yes No Yes No No
Minimal dependencies Yes No Yes Yes Yes No No No No Yes Yes No No No Yes Yes Yes
Embedding 1M nodes (time) ~10s ~30s+GPU ~60s ~10s ~300s ~30s+GPU ~60s+GPU ~120s ~5s+GPU ~300s ~60s ~120s+GPU ~30s ~60s+GPU ~600s ~30s ~20s
Actively maintained Yes Yes Yes Minimal Yes Yes Archived Inactive Inactive Inactive Inactive Inactive Yes Inactive Inactive Inactive Inactive

Summary: Competitive Position

Where pycleora leads:

  1. All-in-one SDK — No other single library combines: embedding + 5 algorithms + classification + community detection + evaluation + 14 datasets + visualization + heterogeneous graphs + streaming + I/O
  2. CPU performance — Rust backend matches or beats pure-Python alternatives by 10-100x
  3. Dynamic graphs — Only library with incremental update, edge removal, streaming, AND inductive embedding
  4. Zero-GPU deployment — Full feature set works on CPU; GPU is optional
  5. Built-in evaluation pipeline — Link prediction, node classification, and clustering metrics without sklearn
  6. Lightweight footprint — ~5 MB install vs 200-600 MB for GPU-dependent competitors

Where competitors lead:

  1. PyG/DGL — 50+ GNN architectures (GCN, GAT, GraphSAGE, etc.) with GPU training
  2. KarateClub — 30+ traditional algorithms (community detection, diffusion, etc.)
  3. PyG/DGL — 80+ real benchmark datasets (not synthetically generated)
  4. PyG/DGL — Active community, papers, tutorials, production deployments
  5. Node2Vec — Simplest API for random-walk based embeddings
  6. GraphVite — Fastest GPU-accelerated embedding (multi-GPU, C++/CUDA core)
  7. StellarGraph — Keras-based API for accessible GNN experimentation (though now archived)
  8. graspologic — Microsoft-maintained spectral methods (ASE, LSE, OMNI) with strong statistical foundations
  9. GraphSAGE — Pioneered inductive graph representation learning

Feature counts:

Category pycleora PyG KarateClub Cleora (Base) Node2Vec DGL StellarGraph GEM GraphVite DeepWalk LINE SDNE graspologic GraphSAGE Struc2Vec VERSE NetSMF
Public functions 52 200+ 40+ 8 ~5 200+ ~80 ~15 ~10 ~5 ~3 ~5 50+ ~10 ~5 ~3 ~3
Algorithms 11 50+ 30+ 1 1 50+ ~15 ~8 ~5 1 1 1 ~5 1 1 1 1
Datasets 14 70+ ~5 0 0 40+ ~10 ~3 ~5 0 0 0 ~5 0 0 0 0
Graph types 6 6 3 2 2 6 5 3 3 1 2 2 3 4 1 3 2
Downstream tasks 7 3 2 0 0 3 4 0 2 0 0 0 0 1 0 0 0
I/O formats 6 3 2 1 1 3 2 1 0 1 0 0 2 0 1 0 0
Dependencies (required) 3 10+ 5+ 0 2 10+ 8+ 5+ 3+ 3 1 5+ 8+ 5+ 3 1 2
Install size ~5 MB ~500 MB+ ~15 MB ~3 MB ~2 MB ~400 MB+ ~600 MB+ ~50 MB ~200 MB+ ~5 MB ~5 MB ~300 MB+ ~50 MB ~500 MB+ ~5 MB ~5 MB ~10 MB