adobe
AutoML Module Design
Adaptive neural architecture learning for neural compression and transfer learning
Problem
The need for efficient neural compression and transfer learning models is becoming increasingly important with the growing size of deep learning models. However, manually designing these models is a time-consuming and challenging process that often requires a significant amount of expertise and resources.
Solution
This project tackled the issue by designing AutoML modules for learning task-adaptive neural architectures. These modules enabled neural compression and transfer learning at both a feature-level and a structure-level, making the process more efficient and cost-effective.
Project details
- Auto ML
- Data integration
- Transfer learning
- Neural networks
32 weeks
U.S
2022-23
