Bayesian Tensor Toolbox (BayTT)

(the repo is still under construction, some link and statistic could be wrong, we will release the full code soon)

BayTT is an open-source library collecting state-of-art models and baselines for Bayesian Tensor decomposition.

We provide a neat code base to decompose a sparse tensor in probabilistic ways, which cover three mainstream tasks now: Sparse Tensor Decomposition, Streaming Tensor Decomposition, Temporal Tensor Decomposition . We will add more topics like Functional Tensor Decomposition in the future.

For each task, we made the leader borad evaluated on several classical datasets. We also provide the dataset in the repo.

Leaderboard

Note: We will keep updating this leaderboard. If you have proposed advanced and awesome models, you can send us your paper/code link or raise a pull request. We will add them to this repo and update the leaderboard as soon as possible.

Compared models of this leaderboard. ☑ means that their codes have already been included in this repo.

Sparse Tensor Decomposition

Model name	Movielens 10K	Movielens 1M	ACC	DBLP
🥇 1st	NEST	NEST	NEST	BASS-Tucker
🥈 2nd	POND	POND	POND	POND
🥉 3rd	SparseHGP	SparseHGP	SparseHGP	SparseHGP

NEST - “Nonparametric Decomposition of Sparse Tensors”, [(ICML 2021)] [Code].
POND - “Probabilistic Neural-Kernel Tensor Decomposition”, [ICDM 2020] [Code].
SparseHGP - “Nonparametric Sparse Tensor Factorization with Hierarchical Gamma Processes”, [ICML 2022] [Code]

Streaming Tensor Decomposition

Model name	Movie-lens	DBLP	ACC	demo
🥇 1st	SFTL	SFTL	SFTL	SFTL
🥈 2nd	SBDT	SBDT	SBDT	SBDT
🥉 3rd	BASS-Tucker	BASS-Tucker	BASS-Tucker	BASS-Tucker

BASS-Tucker - Shikai Fang, Akil Narayan, Robert Kirby, and Shandian Zhe, “Bayesian Continuous-Time Tucker Decomposition ”, The 39 International Conference on Machine Learning [(ICML 2022)] [Code].
SBDT - Shikai Fang, Zheng Wang, Zhimeng Pan, Ji Liu, and Shandian Zhe, “Streaming Bayesian Deep Tensor Factorization” [ICML 2021] [Code].
SFTL - Streaming Factor Trajectory Learning for Temporal Tensor Decomposition [NeurIPS 2023] [Code]

Temporal Tensor Decomposition

Model name	Movie-lens	DBLP	ACC	demo
🥇 1st	SFTL	SFTL	DEMOTE	SFTL
🥈 2nd	NON-FAT	DEMOTE	NON-FAT	DEMOTE
🥉 3rd	BCTT	BCTT	NON-FAT	NON-FAT

DEMOTE - Zheng Wang, Shikai Fang, Shibo Li, and Shandian Zhe, “Dynamic Tensor Decomposition via Neural Diffusion-Reaction Processes” [(NeurIPS 2023)] [Code].
NON-FAT - Zheng Wang, and Shandian Zhe, “Nonparametric Factor Trajectory Learning for Dynamic Tensor Decomposition”[ICML 2022] [Code].
BCTT - Shikai Fang, Akil Narayan, Robert M. Kirby, and Shandian Zhe, “Bayesian Continuous-Time Tucker Decomposition”[ICML 2022] [Code]

List of Byesian Tensor Models in this Repo

Bayesian Sparse Tensor Decomposition

Name	Description
SparseHGP	Sparse Tensor Factorization with Hierarchical Gamma Processes	demo	paper	origin code
NEST	Non-linear decompostion based on Dirichlet processes and Gaussian processes	demo	paper	origin code
POND	Non-linear decompostion based on Deep Kernel Gaussian Process	demo	paper	origin code
GPTF	Non-linear decompostion based on Sparse Gaussian Process	demo	paper	origin code
SVI-CP	Bayesian CP decompostion with stochastic variational inference(SVI)	demo	paper	origin code
SVI-Tucker	Bayesian Tucker decompostion with stochastic variational inference(SVI)	demo	paper	origin code
CEP-CP	Bayesian CP decompostion with conditional expectation propagation(CEP)	demo	paper	origin code
CEP-Tucker	Bayesian Tucker decompostion with conditional expectation propagation(CEP)	demo	paper	origin code

Bayesian Streaming Tensor Decomposition

Name	Description
SNBDT	Streaming Nonlinear decomposition with random Fourier features	demo	paper	origin code
SBDT	Streaming Deep decompostion with sparse BNN	demo	paper	origin code
BASS-Tucker	Streaming Tucker decompostion with sparse Tucker core	demo	paper	origin code
POST	Streaming CP decompostion with SVB update	demo	paper	origin code
ADF-CP	Streaming CP decompostion with ADF update	demo	paper	origin code
ADF-Tucker	Streaming Tucker decompostion with ADF update	demo	paper	origin code

Bayesian Temporal Tensor Decomposition

Name	Description
SFTL	Streaming Temporal CP/Tucker with time-varing latent factors	demo	paper	origin code
DEMOTE	Temporal tensor as Diffusion-Reaction Processes on Graph	demo	paper	origin code
BCTT	Temporal Tucker decompostion with time-varing tucker core	demo	paper	origin code
NON-FAT	GP priors + Fourier Transform	demo	paper	origin code
THIS-ODE	Temporal Tensor decompostion with neuralODE	demo	paper	origin code
CT-GPTF	Streaming CP decompostion with ADF update	demo	paper	origin code
CT-CP	Streaming Tucker decompostion with ADF update	demo	paper	origin code

Usage

Install Python 3.8. For convenience, execute the following command.

pip install -r requirements.txt

Prepare Data. You can obtained the well pre-processed datasets from [Google Drive], [Tsinghua Cloud] or [Baidu Drive]. Then place the downloaded data under the folder ./dataset. Here is a summary of supported datasets.

Citation

If you find this repo useful, please cite our paper.

@inproceedings{fang2022bayesian,
  title={Bayesian Continuous-Time Tucker Decomposition},
  author={Fang, Shikai and Narayan, Akil and Kirby, Robert and Zhe, Shandian},
  booktitle={International Conference on Machine Learning},
  pages={6235--6245},
  year={2022},
  organization={PMLR}
}

xuangu-fang/Bayesian-Tensor-Toolbox