Latent reasoning is a versatile and powerful method for problem solving where computations are adaptively unrolled along depth at test-time. However, current models for latent reasoning are computationally expensive, memory-inefficient and difficult to train. This project will investigate principles to improve on these challenges, namely through event-driven methods and control theory for dedicated implementations of latent reasoning. These investigations will be grounded in hardware through co-design of neuromorphic hardware to improve accuracy, latency, data efficiency, and energy consumption on complex algorithmic or cognitive tasks. This project is part of the ELEVATE MSCA Doctoral Network (https://www.elevate-dn.eu/) and co-supervised by our partners at the university of Liège.
Your tasks in detail:
We work on the very latest issues that impact our society and are offering you the chance to actively help in shaping the change! We support you in your work with:
In addition to exciting tasks and a collegial working environment, we offer you much more: https://go.fzj.de/benefits
We welcome applications from people with diverse backgrounds, e.g. in terms of age, gender, disability, sexual orientation / identity, and social, ethnic and religious origin. A diverse and inclusive working environment with equal opportunities in which everyone can realize their potential is important to us.
The following links provide further information on diversity and equal opportunities: https://go.fzj.de/equality and on specific support options for women: https://go.fzj.de/womens-job-journey