1 min read

Inroads into EDA using Machine Learning

Machine Learning (ML) is the rage these days and we were not untouched by it. Being   immersed in Specifications and dealing with customer queries about register and   sequence specification on a daily basis, we thought, why not help the customer using   Machine Learning. So we decided to use these Machine Learning algorithms.

The task of using Machine Learning was interesting but challenging. We deal with   hardware specifications which are usually written in English. There is also heavy use   of  semiconductor and electronics industry jargons. We had to use NLP (Natural   Language Processing) libraries for parsing and tokenizing the text after much pre-   processing. There are several libraries for ML already in place.

We chose the TensorFlow library because that is becoming the de facto standard. Having incorporated the ML algorithms in IDesignSpec, we are looking for even wider application in all our products. It has been an enlightening experience for us. It is surprising that the ML industry is already so mature and advanced.

We feel that we are late to the ML party, but I think we are early in terms of EDA industry adopting ML in a big way. Agnisys is making headway with home-grown machine-learning algorithms — and applying it to IDS NextGen  multi-platform product which helps user to create SoC specification at an enterprise level.

ic designer's guide to automating design through implementation of semiconductors