https://github.com/g0bel1n/time-series-representation-learning

https://github.com/g0bel1n/time-series-representation-learning

Last synced: about 1 month ago
JSON representation

• Host: GitHub
• URL: https://github.com/g0bel1n/time-series-representation-learning
• Owner: g0bel1n
• Created: 2022-12-23T10:41:45.000Z (about 2 years ago)
• Default Branch: master
• Last Pushed: 2023-01-07T23:42:15.000Z (almost 2 years ago)
• Last Synced: 2024-10-20T05:34:34.354Z (2 months ago)
• Language: Jupyter Notebook
• Size: 769 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
Time Series and representation learning : CNN vs Transformers

Code for PatchTST is from https://github.com/yuqinie98/PatchTST.

| Trainings                | pretrain on | finetuned on | done |

| ------------------------ | ----------- | ------------ | ---- |

| baseline Transformer     | ettm1       | ettm1        |      |

| financial TS Transformer | ive         | ive          |      |

### ToDo List 

- Train Patcht TST on 

 - [ ] ettm1, ive and GunPoint

  - [ ] IVE

  - [ ] Gunpoint

- Train a MCNN on:

  - [ ] ettm1, ive and GunPoint

  - [ ] IVE

  - [ ] Gunpoint

- Look at for GunPoint

  - [ ] CAM

  - [ ] MDS

  - [ ] Attention Map 

- [ ] Same for IVE-classification

### Command lines

for PachTST

```

python patchtst_pretrain.py --dset_pretrain ettm1 --mask_ratio .4 --stride 8 --patch_len 16 --context_points 336 --n_epochs_pretrain 100 --batch_size 128

````

For resnet

```

python resnet_train.py --dset ettm1  --context_points 336 --n_epochs 100 --batch_size 256 --lr 0.01

```

```

python resnet_train.py --dset gunpoint --batch_size 8  --head_type classification  --context_points 150 --target_points 2  --revin 0 --n_epochs 20

```

```

python patchtst_supervised.py --dset gunpoint --batch_size 8 --patch_len 16 --stride 8 --head_type classification --features U --context_points 150 --target_points=2 --revin 0 --n_epochs 30  --is_train 1

```

```

python patchtst_finetune.py --dset_finetune ettm1 --patch_len 16 --stride 8 --batch_size 256 --context_points 336  --features M --target_points 2 --pretrained_model saved_models/ettm1/patchtst/patchtst_pretrained_cw336_patch16_stride8_epochs-pretrain100_mask0.4_model1.pth --is_finetune 1 --revin 0 --head_type classification --classification .05

```

```

python resnet_train.py --dset gunpoint  --batch_size 32  --head_type classification  --context_points 150 --target_points 2  --revin 0 --n_epochs 20 --classification .05

```