Models

Univariate Statistical Learning Models

OU
ARIMA (might AR)

Univariate Machine Learning Models

MLP
MLP + MCCR Loss
Bahdanau Attention
Bahdanau Attention + MCCR

loss function input for Train/Valid/Test set

Train/Valid : Loss between y_hat and y (transformed))
Test : Loss between y_hat_inv (inverse transformed) and y_raw

Multivariate Machine Learning Models

XGBoost
MLP
MLP + MCCR
LSTNet (Skip Layer)
LSTNet (Skip Layer) + MCCR
Transformer
Transformer + MCCR

loss function input for Train/Valid/Test set

Train/Valid : Loss between y_hat and y (transformed))
Test : Loss between y_hat_inv (inverse transformed) and y_raw

Frameworks

PyTorch (>= 1.8.0)
Optuna (>= 2.3.0)
scikit-learn (>= 0.24.0)

Container

Run in Singularity (>= 3.6) container generated torch-py3.def
Run singularity with following commands
- bind paths (/input for input, /mnt/data for output)
- enable nvidia devices by --nv
- containers refer requirements.txt generated from following command
```
$ poetry export -f requirements.txt --output requirements.txt --without-hashes
```

Running models in container (singularity) with commands

TORCH_IMG=my_torch_img.sif
CASES=rnn_mul_lstnet_skip_mccr
CASE_NAME=210818_LSTNet_MCCR
mkdir -p /data/appleparan/"${CASE_NAME}"
singularity exec --nv --bind "${HOME}"/input:/input:ro,/data/"${CASE_NAME}":/mnt/data:rw ${TORCH_IMG} python3 -m mise --dl ${CASES}

Reference

LSTNet: arxiv link
- Lai, Guokun, et al. "Modeling long-and short-term temporal patterns with deep neural networks." The 41st International ACM SIGIR Conference on Research & Development in Information Retrieval. 2018.
Transformer: arxiv link
- Zerveas, George, et al. "A Transformer-based Framework for Multivariate Time Series Representation Learning." arXiv preprint arXiv:2010.02803 (2020).
MCCR (Correntropy based Loss): JLMR link
- Feng, Yunlong, et al. "Learning with the maximum correntropy criterion induced losses for regression." J. Mach. Learn. Res. 16.1 (2015): 993-1034.

Publication

Jongsu Kim and Changhoon Lee. "Deep Particulate Matter Forecasting Model Using Correntropy-Induced Loss." arXiv preprint arXiv:2106.03032 (2021). link
- Accepted in Journal of Mechanical Science and Technology

lusccc / mise.py Goto Github PK

mise.py's Introduction

Models

Univariate Statistical Learning Models

Univariate Machine Learning Models

loss function input for Train/Valid/Test set

Multivariate Machine Learning Models

loss function input for Train/Valid/Test set

Frameworks

Container

Reference

Publication

mise.py's People

Contributors

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent