Lee Center for Advanced Networking
MICHELLE EFFROS

In today’s networked world, the speed at which information travels is a key problem, says Lee Center member Michelle Effros, Associate Professor of Electrical Engineering at Caltech. One way to transmit information faster and more efficiently is to “compress” the data through the use of software algorithms. Algorithms search for and eliminate redundancies within information, reducing the number of bits sent, a process called compression. When the data arrives at the receiving end, other algorithms replace the missing bits, allowing the data to be read and understood.

The usual way to compress data is through the use of an independent code for every sender and receiver in a network. But compressing data this way is inefficient, says Effros. Working with her fellow Lee Center members Robert McEliece, the Allen E. Puckett Professor and Professor of Electrical Engineering, and Babak Hassibi, Assistant Professor of Electrical Engineering, Effros is developing software that can compress data from one source or many sources, whether the data is being sent to one receiver or many.

Effros is also developing “multi-resolution,” or “progressive transmission,” source coding. A single user needs to be able to communicate the same information to a wide array of users at once. But users’ Internet connections and computers often vary in speed and performance and cannot all cope with the same type of coding. Effros’s multi-resolution codes allow an individual to retrieve as much or as little of a file as desired. For example, the recipient of an image file can choose the resolution based on data handling ability.

In 2002, MIT’s Technology Review cited Effros for her innovations in network data compression. The magazine stated that her “heady work has almost single-handedly created research interest in algorithms to optimize transmission of data over busy, noisy networks like the Internet and various wireless infrastructures.” Her algorithms, said the article, have set the “standard” for academic- and commercial-network compression techniques.