Personalized Machine Learning
MAS.S61 Spring 2017, Thursdays 1-3pm, room: E15-359
Instructors
Instructors
Teaching Assistants
Teaching Assistants
Summary
Summary
Nowadays, there is a need more than ever for personalized machine learning models that can learn efficiently from big data and handle challenges of emerging technologies as in smart healthcare and medicine. By personalized we mean the models that are expected to work well for each and every person, and not only the average population - as traditionally approached in machine learning. The Personalized Machine Learning course is designed to equip students with the tools and knowledge necessary for personalized human data analysis, with the focus on health and well-being. This course will provide students with an overview of the cutting-edge machine learning approaches (including Active Learning and Domain Adaptation techniques, with the focus on Deep Neural Networks and Gaussian Processes as modeling tools) that they can later use to build their own creative personalized applications. With the guidance of the instructors, you will learn how to use and manipulate these machine learning models to take the full benefits of the personalized learning in your projects. Each student (or a group) will be provided with a number of datasets, access to machine learning tools, and is expected to actively participate in group discussions in the class - leading to the identification and solving of the key challenges in personalized machine learning. Students are also expected to identify/design/implement their own project (with a guidance of the instructors), where they will get in-depth understanding of the personalized machine learning algorithms. Note that the class is originally intended for the Media Lab students, however, other interested students are also welcome!
Nowadays, there is a need more than ever for personalized machine learning models that can learn efficiently from big data and handle challenges of emerging technologies as in smart healthcare and medicine. By personalized we mean the models that are expected to work well for each and every person, and not only the average population - as traditionally approached in machine learning. The Personalized Machine Learning course is designed to equip students with the tools and knowledge necessary for personalized human data analysis, with the focus on health and well-being. This course will provide students with an overview of the cutting-edge machine learning approaches (including Active Learning and Domain Adaptation techniques, with the focus on Deep Neural Networks and Gaussian Processes as modeling tools) that they can later use to build their own creative personalized applications. With the guidance of the instructors, you will learn how to use and manipulate these machine learning models to take the full benefits of the personalized learning in your projects. Each student (or a group) will be provided with a number of datasets, access to machine learning tools, and is expected to actively participate in group discussions in the class - leading to the identification and solving of the key challenges in personalized machine learning. Students are also expected to identify/design/implement their own project (with a guidance of the instructors), where they will get in-depth understanding of the personalized machine learning algorithms. Note that the class is originally intended for the Media Lab students, however, other interested students are also welcome!
Prerequisites
Prerequisites
A background in linear algebra, probability, ideally advanced statistics (with calculus) and machine learning. Prior experience with using Matlab and/or Python (needed for the final project implementation as well as homework). The groups can be formed so that there is a complimentary knowledge (machine learning, programming and creativity, as we also put a high emphasis on the creativity part -- necessary for identifying novel challenges and approaches to personalized modeling. Depending on your background, we will try to *personalize* the course so that it best fits your interests and the interests of the group you will be doing the final project with! For all the questions, just get in touch via email orudovic@media.mit.edu .
A background in linear algebra, probability, ideally advanced statistics (with calculus) and machine learning. Prior experience with using Matlab and/or Python (needed for the final project implementation as well as homework). The groups can be formed so that there is a complimentary knowledge (machine learning, programming and creativity, as we also put a high emphasis on the creativity part -- necessary for identifying novel challenges and approaches to personalized modeling. Depending on your background, we will try to *personalize* the course so that it best fits your interests and the interests of the group you will be doing the final project with! For all the questions, just get in touch via email orudovic@media.mit.edu .
Enrollment
Enrollment
The class is limited to 20 students taking it for credits. To register for the class, please fill in the online form:
The class is limited to 20 students taking it for credits. To register for the class, please fill in the online form:
*Closed -- the class is already full! Unfortunately, there will be no places for listeners.*
*Closed -- the class is already full! Unfortunately, there will be no places for listeners.*
Exams
Exams
There will be no exams. There will be weekly reading an presentation assignments and a final project. All students are involved in two weeks of presenting work - in one week as the presenter/supporter, and in one week as the critic/attacker. Thus they need to learn to take both sides in understanding the work.
There will be no exams. There will be weekly reading an presentation assignments and a final project. All students are involved in two weeks of presenting work - in one week as the presenter/supporter, and in one week as the critic/attacker. Thus they need to learn to take both sides in understanding the work.
Credits
Credits
The course is structured as follows: 2 hrs of classes on Thursday (students are expected to participate actively in the class. There will be an extra hour on Fridays for recitation of the course materials. This will give a chance to get your 'hands-on' on the machine learning tools that you will be using in your projects. We envision that you will spend around 9 hrs per week working on the course assignments and the final project.
The course is structured as follows: 2 hrs of classes on Thursday (students are expected to participate actively in the class. There will be an extra hour on Fridays for recitation of the course materials. This will give a chance to get your 'hands-on' on the machine learning tools that you will be using in your projects. We envision that you will spend around 9 hrs per week working on the course assignments and the final project.