Graph Algorithms and Machine Learning

Graphs have long been a fundamental way to model relationships in data across industries as diverse as IT, finance, transportation, telecommunications, and cybersecurity. Today, they are increasingly used in machine learning pipelines—enabling clustering for classification tasks, improving recommendation systems, ranking search results, and more. But as the sheer quantity of collected data has grown, so has the complexity of mapping these connections.   

As a result, the efficient processing of large graphs has attracted significant attention, due to its applications in various domains, including social network analysis, epidemiology, computational biology, machine learning, and scientific simulations. Today, graphs have become extremely large and are evolving rapidly in real-time — which has made designing graph analytics a major challenge.

This accelerated course provides a comprehensive overview of critical topics in graph analytics, including applications of graphs, the structure of real-world graphs, fast graph algorithms, synthetic graph generation, performance optimizations, programming frameworks, and learning on graphs. The curriculum additionally covers software performance engineering concepts, such as parallelism, caching, and compression, in the context of graph processing, as well as different design choices that will enable you to use or design the appropriate graph solutions for your needs. 

Through tutorials, exercises, and demonstrations featuring state-of-the-art graph analytics tools, you will broaden your fundamental understanding of graph analytics, and master the techniques and tools that you need to efficiently solve large-scale graph problems in your organization.

Certificate of Completion from MIT Professional Education

Links & Resources

Modeling relationships to solve complex problems efficiently, MIT News, October 4, 2024

By participating in this course, you will:

• Learn how to model structured data with graphs
• Enhance your understanding of real-world graph properties and how to generate synthetic graphs
• Master fundamental graph algorithms
• Discuss parallelism and how it can be used to speed up graph processing
• Examine performance characteristics of graph algorithms
• Assess the state-of-the-art graph processing tools available today and learn to use certain graph software
• Explore the pros and cons of different graph processing approaches
• Acquire a new set of tools for improving the effectiveness and performance of machine learning pipelines

Course Outline

This course runs 9:30 am - 2:00 pm EDT each day. 

Schedule

Day 1

• 9:30-10:30 am: Introduction to Graph Theory and Applications of Graphs
  Review the basics of graph theory and its application.
• 10:30-10:45 am: Break
• 10:45 am-11:30 pm: Structure of Real-World Graphs 
  Study properties of real-world graphs.
• 11:30am-12:30 pm: Lunch Break
• 12:30-1:30 pm: Structure of Real-World Graphs 
  Study how real-world graphs are formed.
• 1:30-2:00 pm: Q&A

Day 2

• 9:30-10:30 am:Graph Algorithms
  Learn about several fundamental graph algorithms.
• 10:30-10:45 am: Break
• 10:45-11:30 am: Graph Algorithms
  Learn about several fundamental graph algorithms.
• 11:30am-12:30pm: Lunch Break
• 12:30-1:30pm: Demo and Exercises with Graph Processing Software (NetworkX)
  Learn how to use a popular graph processing software.
• 1:30-2:00 pm: Q&A

Day 3

• 9:30-10:30 am: Machine Learning on Graphs
  Study methods for Web search
• 10:30-10:45 am: Break
• 10:45-11:30 am: Machine Learning on Graphs 
  Learn about how to do classification and prediction on graphs.
• 11:30am-12:30 pm: Lunch Break
• 12:30pm-1:30 pm: Machine Learning on Graphs 
  Discuss case studies of using machine learning on graphs.
• 1:30-2:00 pm: Q&A

Day 4

• 9:30-10:45 am: Large-Scale Graph Processing Frameworks
  Learn about frameworks for writing fast code for processing graphs.
• 10:45-11:00 am: Break
• 11:00-11:45 am: Problem Clinic
  We will discuss graph-related problems in small groups.
• 11:45 am-12:45 pm: Lunch Break
• 12:45-1:30 pm: Problem Clinic
  Each group will report to the class.
• 1:30-2:00 pm: Q&A

Links and Resources

• Using Algorithms to Root Out Fraud
  • AI can play a role in creating strong internal controls and fraud prevention, key elements of upholding and embodying ethics.
• The Tenured Engineers of 2024
  • Julian Shun, associate professor of computer science and principal investigator at CSAIL, focuses on the theory and practice of parallel and high-performance computing. 

This course is designed for scientists and engineers in industry or government who work with large-scale data. The strategies covered are applicable to a variety of fields, such as software/IT, finance, transportation, biotech, telecommunications, and cybersecurity.

Applicable roles include, but are not limited to:

• Data scientists who want to improve their ability to extract actionable insights from large and structured data sets
• Software engineers looking to develop fast graph software
• Project managers who wish to increase their efficiency in overseeing technical teams working on graph-related projects and graph software products
• Any technical professionals who want to use data to inform business decisions, accelerate discoveries, detect fraud, defend against security threats, increase revenue, or improve service quality 

Requirements

Participants should have general knowledge of computer science at an undergraduate level, as well as some programming experience—Python and C++ are preferred. The NetworkX Python package will be used for some demonstrations and exercises, and so your laptop or computer must be compatible with the software.