The computing sector accounted for an estimated three percent of global CO2 emissions in 2022 — more than Spain, Italy, France, and Portugal combined. Still, compared to the staggering impacts of... Read more about Udit Gupta leads $2 million project to tackle tech’s environmental toll
Welcome Jayadev Acharya
- New Faculty Year: 2016
Jayadev Acharya joined the faculty of Cornell Engineering in July, 2016. Acharya is an assistant professor in the School of Electrical and Computer Engineering (ECE) and a graduate field member in Computer Science (CS). Acharya’s research interests are Information Theory, Algorithmic Statistics, and Machine Learning. The focus of his work is understanding and achieving the fundamental tradeoffs between resources that go into a machine learning problem.
“You don’t want to ask SIRI a question and then have to wait a day for the perfect answer,” says Acharya. Acharya’s work has far broader applications than simply how quickly SIRI tells you the closest location where you can buy a new snow shovel. Machine learning algorithms are becoming central to many of the tasks individuals, businesses, and governments perform on desktops, laptops, smartphones, and tablets each day. To perform the job they are tasked with, machine learning algorithms are constrained by resources such as the amount of training data available, the memory (space) on the device, and the time required for implementation of the algorithm. Each of the factors involved can be adjusted.
Acharya designs mathematical models to understand the trade-offs between these, and other resources. The goals are two-fold: to establish the information theoretic limits of the resource-trade-offs, and strive to achieve them. If a device or a system has unlimited storage, unlimited data, unlimited computation power, then these trade-offs might not be necessary. But if, for example, your smartphone has limited memory available and the app you are using requires lots of locally-stored data, this will have a negative effect on the overall performance of the phone. “Surprisingly,” says Acharya, “these trade-offs have not yet been studied very closely.”
Acharya grew up in the Indian state of Odisha. He found himself more interested in math than in other subjects as an elementary and secondary school student. He earned his Bachelor’s Degree in Electronics and Communication Engineering from the Indian Institute of Technology (IIT), Kharagpur. He then moved to San Diego, where he earned both his Master’s and his Ph.D. in Electrical and Computer Engineering from the University of California, San Diego (UCSD). Acharya then spent two years in a postdoctoral researcher position in MIT’s Theory of Computation Group before joining Cornell Engineering in 2016.
Acharya’s advisor at UCSD was Professor Alon Orlitsky, who is the Qualcomm Chair for Information Theory and Its Applications. “During our meetings, Professor Orlitsky used to ask for the simplest examples of problems that we do not know how to solve. Solving that special case often paved the path for solving the general case,” says Acharya. While at UCSD, Acharya focused on probability estimation and compression over “large domains”, where the set of possible observations could be much larger than the amount of data we have seen. “The setting is not very different from what I. J. Good and Alan Turing did to break the Enigma machine during WWII. They had to find the distribution of secret keys, many of which were not used before.” One of Acharya’s papers written during his Ph.D. studies provides the fundamental limits and algorithms for the “Good-Turing probability estimation.”
In his postdoctoral position, Acharya studied various statistical problems in the large domain settings. He designed algorithms for measuring randomness in data, and testing whether data sources have certain properties of interest. The primary focus in these applications was understanding the trade-off between data and accuracy. Acharya will continue this work and explore the trade-offs between data, memory, time, and other factors for problems in statistical learning now that he is in Ithaca.
“I am excited to be at Cornell. It is strong in ECE, Operations Research (OR), and Computing and Information Science (CIS), fields I have keen interest in. The avenues for collaboration are plenty!,” says Acharya enthusiastically. In the spring semester of 2017, Acharya will be teaching a class on machine learning geared especially to ECE majors.