Semester.ly

Johns Hopkins University | EN.600.476

Machine Learning: Data to Models

3.0

credits

Average Course Rating

(3.55)

How can robots localize themselves in an environment when navigating? Can we predict which patients are at greatest-risk for complications in the hospital? Which movie should I recommend to this user given his history of likes? Many such big data questions can be answered using the paradigm of probabilistic models in machine learning. These are especially useful when common off-the-shelf algorithms such as support vector machines and k-means fail. You will learn methods for clustering, classification, structured prediction, recommendation and inference. We will use Murphy's book, Machine Learning: a Probabilistic Perspective, as the text for this course. Assignments are solved in groups of size 1-3 students. The class will have 4 interactive sessions during which we brainstorm how to solve example open-ended real-world problems with the tools learnt in class. Students are also required to do a project of their choice within which they experiment with the ideas learnt in class. [Analysis or Applications] Students may receive credit for EN.600.476 or EN.600.676, but not both. Requistes include Intro Prob/Stat, Linear Algebra and Intro Machine Learning as well as strong background in s.

Fall 2013

(3.89)

Spring 2013

(3.1)

Spring 2015

(3.67)

Fall 2013

Professor: Suchi Saria

(3.89)

Students thought that the good aspects of this course included the hands-on, practical application of the materials. They liked choosing their own projects and liked the breadth of topics covered over the course. They thought that the course was somewhat unorganized and that the course did not always stick to the syl abus. Suggestions for improvement included providing more lectures and additional assignments to support the final project. Prospective students should have an idea for their final project before taking the course and know the basics to researching and sharing academic papers. 88

Spring 2013

Professor: Suchi Saria

(3.1)

The best aspects of this course included the coverage of a brand new field of computer science, and that the material introduced was intriguing. Some students felt the class was often unstructured. One suggestion was to focus the curriculum to learn fewer things but in more depth throughout the duration of the semester. Another suggestion was to implement more structure. Prospective students are encouraged to come into this class with a working knowledge of NLP, Vision, AI, and ML with Dredze.

Spring 2015

Professor: Suchi Saria

(3.67)

The best aspects of the class included the interesting subject matter, the entertaining and interactive lectures, and the exposure to cutting edge developments in the field. Delayed feedback on assignments prevented students from having opportunities to improve, and the exceptional y fast pace of the class made it difficult to keep up with concepts. As a result, students suggested slowing the pace in class and giving shorter assignments that emphasized understanding rather than task completion. Prospective students will benefit from a background in linear algebra, machine learning, and statistics. This course was based more in programming than theory, and dealt largely with statistical modeling.