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 (jobin.wilson@flytxt.com)
2.Amit Kumar Meher (amit.meher@flytxt.com)
3.Bivin Vinodkumar Bindu (bivin.vinod@flytxt.com)
4.Manoj Sharma (mksnith@gmail.com)
5.Vishakha Pareek (vishakhapareek@ceeri.res.in)
6.Prof.Santanu Chaudhury
7.Prof.Brejesh Lall
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} }