Topics in Machine Learning Systems
The course will cover recent papers from the literature in the emerging area of ML systems. With the emergence of massive data and data science, machine learning is widely applicable in a variety of usage scenarios with high performance, accuracy and cost being key design goals. The latter not only has implications for algorithms but also platforms from software to hardware to enable collective optimization of the design metrics. The topic is inherently multidisciplinary and will cover papers from a variety of conferences in computer science subfields.
Students will understand the state-of-the-art in the emerging area of ML Systems. Specifically, the course will cover core technologies in production ML systems including:
- languages and paradigms for specification of large-scaling machine learning applications,
- the convergence of analytics from relational databases to unstructured data,
- resource management in large-scaling ML systems,
- network stacks for ML systems,
- emerging ML systems accelerator architecture.
Who should take CS 723?
CS 723 is a graduate course and is highly recommended for master and PhD students. Like other graduate-level courses, the course includes weekly readings, discussions, and questions on papers of seminal and recent contributions to the field of systems for machine learning.
Readings and Presentations
In this course, we will read papers, and take turn presenting them. It is absolutely important to read
the papers prior to attending class because the class will proceed in the form of a discussion among
participants and a presentation to introduce the main topics covered in the papers. The students will
take turn presenting throughout the semester.
Attendance
You are expected to be in the classroom and actively participate in the discussions.
Prerequisites
Graduate level course in computer architecture, programming languages, and/or systems
Submission System
Submit your write-ups to Martijn de Vos (
martijn.devos@epfl.ch) before Monday noon.
Grading
The students will be graded based on class discussions, presentations and short reviews written for each reading assignment.
