https://github.com/flytxtds/AutoGBT

AutoGBT is used for AutoML in a lifelong machine learning setting to classify large volume high cardinality data streams under concept-drift. AutoGBT was developed by a joint team ('autodidact.ai') from Flytxt, Indian Institute of Technology Delhi and CSIR-CEERI as a part of NIPS 2018 AutoML for Lifelong Machine Learning Challenge.
https://github.com/flytxtds/AutoGBT

automl concept-drift hyperopt lifelong-ml nips-2018

Last synced: 3 days ago
JSON representation

AutoGBT is used for AutoML in a lifelong machine learning setting to classify large volume high cardinality data streams under concept-drift. AutoGBT was developed by a joint team ('autodidact.ai') from Flytxt, Indian Institute of Technology Delhi and CSIR-CEERI as a part of NIPS 2018 AutoML for Lifelong Machine Learning Challenge.

• Host: GitHub
• URL: https://github.com/flytxtds/AutoGBT
• Owner: flytxtds
• License: other
• Created: 2018-12-01T13:27:19.000Z (almost 6 years ago)
• Default Branch: master
• Last Pushed: 2019-12-13T05:19:07.000Z (almost 5 years ago)
• Last Synced: 2024-08-02T07:13:52.490Z (3 months ago)
• Topics: automl, concept-drift, hyperopt, lifelong-ml, nips-2018
• Language: Python
• Homepage:
• Size: 271 KB
• Stars: 114
• Watchers: 7
• Forks: 41
• Open Issues: 0
• Metadata Files:
  • Readme: readme.md
  • License: COPYING.txt

Awesome Lists containing this project

• automl - Homepage

README

        
# AutoGBT

AutoGBT stands for Automatically Optimized Gradient Boosting Trees, and is used for AutoML in a lifelong machine learning setting to classify large volume high cardinality data streams under concept-drift. AutoGBT was developed by a joint team ('autodidact.ai') from Flytxt, Indian Institute of Technology Delhi and CSIR-CEERI as a part of NIPS 2018 AutoML Challenge (The 3rd AutoML Challenge: AutoML for Lifelong Machine Learning). Our team won the first prize in the challenge. More details of the challenge is available at https://www.4paradigm.com/competition/nips2018. The work will be presented at NIPS 2018 during the Competition Track session (https://nips.cc/Conferences/2018/Schedule?showEvent=10945).

More details are available in our paper: https://link.springer.com/chapter/10.1007/978-3-030-29135-8_13

Team:\

1.Jobin Wilson ([email protected])\

2.Amit Kumar Meher ([email protected])\

3.Bivin Vinodkumar Bindu ([email protected])\

4.Manoj Sharma ([email protected])\

5.Vishakha Pareek ([email protected])\

6.Prof.Santanu Chaudhury\

7.Prof.Brejesh Lall

# How to Run

Download the starter kit from the NIPS AutoML from competion webpage (https://competitions.codalab.org/competitions/20203#participate-get_starting_kit) and setup locally as instructed in the readme file within the starter kit. Copy the folder "AutoGBT" into the starting_k folder inside the starter kit. Install docker from https://docs.docker.com/get-started/ and issue the following command to invoke the docker image corresponding to python3 bundle for the challenge.\

docker run -it -u root -v $(pwd):/app/codalab codalab/codalab-legacy:py3 bash

For ingestion, use the following command from the docker shell prompt

python3 AutoML3_ingestion_program/ingestion.py AutoML3_sample_data AutoML3_sample_predictions AutoML3_sample_ref AutoML3_ingestion_program AutoGBT\

For scoring, use the following command from the docker shell prompt\

python3 AutoML3_scoring_program/score.py 'AutoML3_sample_data/*/' AutoML3_sample_predictions AutoML3_scoring_output

If you used AutoGBT in one of your projects, please consider citing us:

@incollection{wilson2020automatically,

  title={Automatically Optimized Gradient Boosting Trees for Classifying Large Volume High Cardinality Data Streams Under Concept Drift},

  author={Wilson, Jobin and Meher, Amit Kumar and Bindu, Bivin Vinodkumar and Chaudhury, Santanu and Lall, Brejesh and Sharma, Manoj and Pareek, Vishakha},

  booktitle={The NeurIPS'18 Competition},

  pages={317--335},

  year={2020},

  publisher={Springer}

}