search

blobcity / autoai

Python based framework for Automatic AI for Regression and Classification over numerical data. Performs model search, hyper-parameter tuning, and high-quality Jupyter Notebook code generation.
Apache License 2.0
175 stars 43 forks source link

Add HistGradientBoostingClassifier #41

Closed sanketsarang closed 3 years ago

sanketsarang commented 3 years ago

Add HistGradientBoostingClassifier model into the library.

Primary File to Change: https://github.com/blobcity/autoai/blob/main/blobcity/config/classifier_config.py

Reference HistGradientBoostingClassifier Implementation: https://github.com/blobcity/ai-seed/blob/main/Classification/Histogram-Based%20Gradient%20Boosting%20Trees/HistGradientBoostingClassifier.ipynb

Dependencies if any, must be appropriately added. Test run of the train function on a classification problem must pass, and the function must attempt to train an HistGradientBoostingClassifier as a potential best fit model.

Devolta05 commented 3 years ago

Hello, I would like to work on this issue. Could this be assigned to me? I have a few question in regards to the project. I am not able to find any documentation for the same (README works but I would like to understand how it is able to find the correct model to fit for the data?). Do I need to have any prior information about the model implementation or the python notebook shared would suffice?

sanketsarang commented 3 years ago

@Devolta05 issue is now allotted to you.

The project is new and we are still in the process of adding Wikis on the internal working. You can look at the modelSearch function inside this file for a better understanding of how the training actually happens: https://github.com/blobcity/autoai/blob/main/blobcity/main/modelSelection.py

For this task, you are not expected to change the modelSearch or the driver functionality. Just adding the extra configuration in the dictionary, along with adding any required dependencies should suffice.

Devolta05 commented 3 years ago

85 Has been merged with the functionality of this Issue. This can be closed in my opinion :thinking: