Feeding Hungry Models Less: Deep Transfer Learning for Embedded Memory PPA Models : Special Session

Supervised machine learning requires large amounts of labeled data for training. In power, performance and area (PPA) estimation of embedded memories, every new memory compiler version is considered independently of previous versions. Since the data of different memory compilers originate from similar domains, transfer learning may reduce the amount of supervised data required by pre-training PPA estimation neural networks on related domains. We show that provisioning times of PPA models for new compiler versions can be reduced significantly by exploiting similarities across versions and technology nodes. Through transfer learning, we shorten the time to provision PPA models for new compiler versions by 50% to 90%, which speeds up time-critical periods of the design cycle. This is achieved by requiring less than 6,500 ground truth samples for the target compiler to achieve average estimation errors of 0.35% instead of 13,000 samples. Using only 1,300 samples is sufficient to achieve an almost worst-case (98th percentile) error of approximately 3% and allows us to shorten model provisioning times from over 40 days to less than one week.