Useful Research Papers for Applied Machine Learning Practitioners
AI-ML has benefitted from the extensive research on the subject fuelled by the business needs to optimize their processes. An AI-ML practitioner needs to be up to date with the latest research to do justice to his art. In this post, I have listed three papers on popular techniques used by practitioners for applied problems. Happy Reading !!
CatBoost
CatBoost (https://catboost.ai) is a recent simple to use supervised machine-learning algorithm based on gradient boosted decision trees. The algorithm is developed and supported by Yandex N.V. from Russia. The algorithm has gained a lot of popularity because it supports categorical variables out of the box. .
Dorogush, Anna Veronika, Vasily Ershov, and Andrey Gulin. “CatBoost: gradient boosting with categorical features support.” [paper link]
SHAP
Explainability is extremely useful for supervised machine learning to help understand the model predictions and decide how to act upon them. SHAP (SHapley Additive exPlanations) (https://github.com/slundberg/shap) is a popular game theoretic-based approach to identify the contribution of each feature towards the prediction. SHAP provides explainability support for CatBoost model described above.
Lundberg, Scott, and Su-In Lee. “A unified approach to interpreting model predictions.” [paper link]
HyperOpt
HyperOpt (http://hyperopt.github.io/hyperopt/) is a popular package to optimize the hyperparameter tuning of a machine learning model. HyperOpt builds a Bayesian model to understand the impact of change in hyperparameters on model performance. It uses the Bayesian model to optimize the hyperparameter tuning.
Bergstra, James, Daniel Yamins, and David Cox. “Making a science of model search: Hyperparameter optimization in hundreds of dimensions for vision architectures.” [paper link]
