07-17-2026, 04:40 PM
How to find patterns in large graphs
BY Christos Faloutsos
Summary
In his insightful presentation, computer scientist Christos Faloutsos dives into the fascinating world of network analysis, explaining how we can spot hidden patterns and bizarre anomalies inside massive graphs like Facebook, LinkedIn, and phone networks. He reveals that real-world connections aren't random at all; instead, they follow heavy-tailed power law distributions where standard, textbook concepts like "averages" simply break down because human behavior is naturally skewed. By mapping features like overlapping friend groups and tracing the duration of phone calls, his team developed powerful tools—such as the Pegasus system built on Hadoop—capable of scanning billions of data points in mere days.
This structural perspective allows researchers to instantly flag unusual activity, from coordinated adult advertising rings on Twitter and multi-account auction fraud on eBay, to calculating exactly how quickly an epidemic or a fresh idea will ripple through a population. Ultimately, Faloutsos underscores that as data explodes to a petabyte scale, it reveals unique human and computer behaviors that completely defy traditional statistics and simply cannot be seen in smaller samples.
LECTURE
BY Christos Faloutsos
Summary
In his insightful presentation, computer scientist Christos Faloutsos dives into the fascinating world of network analysis, explaining how we can spot hidden patterns and bizarre anomalies inside massive graphs like Facebook, LinkedIn, and phone networks. He reveals that real-world connections aren't random at all; instead, they follow heavy-tailed power law distributions where standard, textbook concepts like "averages" simply break down because human behavior is naturally skewed. By mapping features like overlapping friend groups and tracing the duration of phone calls, his team developed powerful tools—such as the Pegasus system built on Hadoop—capable of scanning billions of data points in mere days.
This structural perspective allows researchers to instantly flag unusual activity, from coordinated adult advertising rings on Twitter and multi-account auction fraud on eBay, to calculating exactly how quickly an epidemic or a fresh idea will ripple through a population. Ultimately, Faloutsos underscores that as data explodes to a petabyte scale, it reveals unique human and computer behaviors that completely defy traditional statistics and simply cannot be seen in smaller samples.
LECTURE
┌────────────────────────────────┐
│ KONSTANTINOS MICHAILIDIS │
└────────────────────────────────┘
│ KONSTANTINOS MICHAILIDIS │
└────────────────────────────────┘