TimeVis

Official source code for IJCAI 2022 Paper: Temporality Spatialization: A Scalable and Faithful Time-Travelling Visualization for Deep Classifier Training

Usage

Download dependencies

Please run the following commands to install all dependencies:

$ pip install -r requirements.txt

Run

1. Prepare data

1.1 Save subject models and data following the following format.

data(input path)
└───Model
│   │   model.py (which contains subject model.)
│   │
│   └───Epoch_1
│       │   index.json (The index list of training data.)
│       │   subject_model.pth (state_dict)
|       |   (train_data.npy) *generated by TimeVis*
|       |   (test_data.npy) *generated by TimeVis*
|       |   (border_centers.npy) *generated by TimeVis*
│   └───Epoch_2
|       |   ...
│   
└───Training_data
|   │   training_dataset_data.pth
|   │   training_dataset_label.pth
│   
└───Testing_data
│   │   testing_dataset_data.pth
│   │   testing_dataset_label.pth

1.2 Set training hyperparameters in ~/TimeVis/singleVis/config.json

2. Train a visualization model

$ python main.py ---content_path /path/to/subject_models --dataset dataset_name -g gpu_id

3. Evaluate visualization model

$ python test.py ---content_path /path/to/subject_models --dataset dataset_name -g gpu_id

Training hyperparameters

Hyperparameters	Meaning	Example
Config Name	The config name for one training process	"cifar10"
NET	The subject model name to be called	"resnet18"
TRAINING_LEN	Training data len	50000
TESTING_LEN	Testing data len	10000
LAMBDA	The trade-off between umap loss and reconstruction loss. It depends on dataset.	10.
L_BOUND	The	.5
MAX_HAUSDORFF	r0
ALPHA	\alpha
BETA	\beta
HIDDEN_LAYER	The number of hidden layers for our visualization model.	3
INIT_NUM
EPOCH_START
EPOCH_END
EPOCH_PERIOD
N_NEIGHBORS	The	15
MAX_EPOCH
S_N_EPOCHS
B_N_EPOCHS
T_N_EPOCHS
PATIENT	Early stopping patient.	3

Reference

If you find our tool helpful, please cite the following paper:

@inproceedings{yang2022temporality,
  title={Temporality Spatialization: A Scalable and Faithful Time-Travelling Visualization for Deep Classifier Training},
  author={Yang, Xianglin and Lin, Yun and Liu, Ruofan and Dong, Jin Song},
  booktitle = {Proceedings of the Thirty-First International Joint Conference on
               Artificial Intelligence, {IJCAI-22}},
  year={2022}
},
@inproceedings{yang2022deepvisualinsight,
  title={DeepVisualInsight: Time-Travelling Visualization for Spatio-Temporal Causality of Deep Classification Training},
  author={Yang, Xianglin and Lin, Yun and Liu, Ruofan and He, Zhenfeng and Wang, Chao and Dong, Jin Song and Mei, Hong},
  booktitle = {The Thirty-Sixth AAAI Conference on Artificial Intelligence (AAAI)},
  year={2022}
}