SLIDE: Smart Algorithms Over Hardware Acceleration for Large-Scale Deep Learning with Beidi Chen - #356
Research#deep learning📝 Blog|Analyzed: Dec 29, 2025 08:04•
Published: Mar 12, 2020 04:43
•1 min read
•Practical AIAnalysis
This article discusses Beidi Chen's work on SLIDE, an algorithmic approach to deep learning that offers a CPU-based alternative to GPU-based systems. The core idea involves re-framing extreme classification as a search problem and leveraging locality-sensitive hashing. The team's findings, presented at NeurIPS 2019, have garnered significant attention, suggesting a potential shift in how large-scale deep learning is approached. The focus on algorithmic innovation over hardware acceleration is a key takeaway.
Key Takeaways
- •SLIDE is a CPU-based algorithmic alternative to GPU-based deep learning.
- •The approach reframes extreme classification as a search problem.
- •Locality-sensitive hashing is a key technique used in SLIDE.
Reference / Citation
View Original"Beidi shares how the team took a new look at deep learning with the case of extreme classification by turning it into a search problem and using locality-sensitive hashing."