IIT-Hyderabad researchers develop method to further understand AI technology

Researchers at Indian Institute of Technology (IIT) Hyderabad have developed a method by which the inner workings of Artificial Intelligence (AI) models can be understood. This would help understand the artificial neural networks or ANN which are AI models and programs that mimic the working of the human brain so that machines can learn to make decisions in a more human-like manner.

“We have proposed a new method to compute the average causal effect of an input neuron on an output neuron. It is important to understand which input parameter is ‘causally’ responsible for a given output," said Vineeth N Balasubramanian, the key researcher on this project at IIT Hyderabad. For example in the field of medicine, how does one know which patient attribute was causally responsible for the heart attack? "Our (IIT Hyderabad research) method provides a tool to analyse such causal effects,” Balasubramanian explained.

The most recent ANNs also known as Deep Learning (DL), help machines train themselves to process and learn from data that has been supplied to them as input, and almost match human performance in many tasks. But one does not know how these machines arrive at decisions, making them less useful when the reason for decisions is necessary, according to a release shared by the institute.

Besides Balasubramanian, his student researchers- Aditya Chattopadhyay, Piyushi Manupriya, and Anirban Sarkar are also part of this research team. Their work has recently been published in the proceedings of 36th International Conference on Machine Learning, one of the highest-rated conferences in AI and ML, the institute stated.

A key bottleneck in accepting such deep learning models in real-life applications, especially risk-sensitive ones, is the ‘interpretability problem.’ The DL models, because of their complexity and multiple layers, become virtual black boxes that cannot be deciphered easily. "Thus, when a problem arises in the running of the DL algorithm, troubleshooting becomes difficult, if not impossible," said Balasubramanian.