This summer I worked with Dr. Jason Moore in the Penn Institute of Biomedical Informatics at the Perelman School of Medicine. I assisted Dr. Moore with his research project PennAI, an open-source, user friendly artificial intelligence (AI) system that selects machine learning methods from an analysis of data. In this research, I designed and programmed a new recommender system to recommend machine learning algorithms based on metrics in the data sets.
PennAI was designed to perform analysis of complex data sets in health care domains. This AI system will be launched within the next year to the entire Penn community. An AI system like this can be used not only for large health datasets, but also can be extended to business research or any type of research that requires analysis of complex data.
My automated machine learning project for the summer was to design a recommender system. I chose to design a system that tested which machine learning algorithms were significantly different from one another. My recommender was human inspired since I also had my recommender select algorithms based on an index that would weight factors that a human might consider when selecting classifiers. I included factors such as interpretability and duration of the run in addition to balanced accuracy.
Through this research, I familiarized myself with software tools that I had never used before including Scikit-learn, a machine learning library in Python, and Pandas Dataframe, a package used for organizing large sets of data. Furthermore, I familiarized myself with machine learning methods and how to code an artificial intelligence system from working with PennAI.
In addition to my work with PennAI, I also worked on creating an interface with PennAI with Amazon Alexa. As part of this work, I researched how to interface Amazon Alexa with an AI system.
This summer I worked with Dr. Jason Moore in the Penn Institute of Biomedical Informatics at the Perelman School of Medicine. I assisted Dr. Moore with his research project PennAI, an open-source, user friendly artificial intelligence (AI) system that selects machine learning methods from an analysis of data. In this research, I designed and programmed a new recommender system to recommend machine learning algorithms based on metrics in the data sets.
PennAI was designed to perform analysis of complex data sets in health care domains. This AI system will be launched within the next year to the entire Penn community. An AI system like this can be used not only for large health datasets, but also can be extended to business research or any type of research that requires analysis of complex data.
My automated machine learning project for the summer was to design a recommender system. I chose to design a system that tested which machine learning algorithms were significantly different from one another. My recommender was human inspired since I also had my recommender select algorithms based on an index that would weight factors that a human might consider when selecting classifiers. I included factors such as interpretability and duration of the run in addition to balanced accuracy.
Through this research, I familiarized myself with software tools that I had never used before including Scikit-learn, a machine learning library in Python, and Pandas Dataframe, a package used for organizing large sets of data. Furthermore, I familiarized myself with machine learning methods and how to code an artificial intelligence system from working with PennAI.
In addition to my work with PennAI, I also worked on creating an interface with PennAI with Amazon Alexa. As part of this work, I researched how to interface Amazon Alexa with an AI system.