Adaptive and secure deep Learning On Heterogeneous Architectures
DL algorithms are designed to improve precision, without considering the limitations of the device that will execute them. Thus, the deployment of DL algorithms on heterogeneous architectures is highly time consuming without adequate support from software development tools.
Moreover, DL applications present bias in their decisions resulting in ethical problems in many companies, big and small.
The objective of ALOHA is to develop a tool flow for helping developers to design, build and deploy DL application on (embedded) heterogeneous architectures. This tool flow will also include security tools to improve resilience of the systems to attacks and secure data in an edge computing paradigm.
From CA’s perspective, this project will help us answer the question of how we can use Agile and DevSecOps to build DL applications while minimizing data bias and supporting fairness and ethical use.