Aaron Wagner, professor of electrical and computer engineering at Cornell, has been thinking about the fundamental assumptions surrounding the use of feedback information in digital communication strategies since he was a graduate student.
Almost every communication system involves some form of feedback. In practice, forward communication links are usually paired with a feedback link from the receiver to the transmitter, facilitating communication in both directions. Several methods for using feedback links in digital communications have been well-established for decades, and although they are common in practice, it is not well understood theoretically how they can be most effectively used.
Wagner wondered if the list was exhaustive. Was there some new method for using feedback that had yet to be uncovered after all these years? Inspiration came from some unlikely sources: ecology and sports.
Animals have been observed to follow more-risky foraging strategies when near starvation and less-risky strategies when food reserves are high. Similar behavior is found in sports like hockey when a team pulls its goalie in order to field an extra offensive player if it is losing late in the game. Risky strategies make more sense when cautious play is guaranteed to lose. Sometimes it’s fine to be timid, but when circumstances change, strategies must become bolder.
“That idea can be applied to communication as well, if it is applied in the right way” Wagner said, “and despite the maturity of the field, it had never been noticed before.” The idea is that the transmitter can track the progress of the communication via the feedback link and switch to a bold signaling strategy when failure looks likely, akin to a hockey team pulling its goalie.
Wagner, working with Nirmal Shende, Ph.D. ’19, and Yücel Altuğ, Ph.D. ’13, described the idea, which they called timid/bold coding, in their paper titled “A New Method for Employing Feedback to Improve Coding Performance,” which was published in the IEEE Transactions on Information Theory and was recently recognized with the 2022 IEEE Information Theory Society Paper Award.
“We found a new way of using the feedback link to improve communication,” Wagner said. “It’s qualitatively different from what had come before, and it provides a new way of looking at a question that was viewed as settled a long time ago.”
In the context of feedback communication, timid/bold coding improves achievable data rates compared with the best non-feedback schemes for certain kinds of channels. The team showed that for a large class of channels, feedback is at all useful only if timid/bold coding is applicable. “So it is unlikely that there is yet another mechanism for using feedback lurking there, waiting to be uncovered,” Wagner said.
Researchers and engineers now have one more proven method for using feedback links to improve their designs. For his part, Wagner found himself approaching personal and professional decisions differently after working on this project. “Play timid when you are expecting to win and play bold when you expecting to lose – it’s a useful rule for many aspects of life.”