Link: https://eu02web.zoom-x.de/j/66719911736?pwd=N5n96kFJbauLi2u79eJI0ZD15hgNsi.1
4:30pm Central European time is (usually) 7:30am Pacific time and 11:30pm Beijing time
Advances in AI today are undeniably impressive, yet modern AI systems often falter when faced with out-of-distribution (OOD) data and hence lack generalization. A promising solution lies in integrating neural nets with symbolic computing in a framework that ensures scalable and sustainable computing. To address this need, we introduce the neuro-vector-symbolic architecture (NVSA), which seamlessly combines deep neural nets with vector-symbolic architectures. NVSA empowers neural nets with compositionally structured representations and operators, enabling robust perception and reasoning, even when confronted with OOD data. Crucially, NVSA’s efficient implementation is informed and benefitted by unique physical properties of in-memory computing hardware, such as constant-time vector-matrix multiplication, in-situ progressive crystallization, and the inherent stochasticity of phase-change memory devices. This synergy paves the way for AI systems that reliably handle learning and reasoning tasks with unprecedented computational and sample efficiency.
Abbas Rahimi received the B.S. degree in computer engineering from the University of Tehran in 2010, and the M.S. and Ph.D. degrees in computer science and engineering from the University of California San Diego in 2015, and subsequently was a postdoctoral fellow at the University of California Berkeley and the ETH Zürich. In 2020, he joined the IBM Research-Zürich laboratory as a Research Staff Member.
H e has received the 2015 Outstanding Dissertation Award in the area of "New Directions in Embedded System Design and Embedded Software" from the European Design and Automation Association, and the ETH Zürich Postdoctoral Fellowship in 2017. He was a co-recipient of the Best Paper Nominations at DAC (2013) and DATE (2019), and the Best Paper Awards at BICT (2017), BioCAS (2018), and IBM's Pat Goldberg Memorial (2020).
