Improved Dropout for Shallow and Deep Learning
Zhe Li, Boqing Gong, Tianbao Yang
–Neural Information Processing Systems
However, the independent sampling for dropout could be suboptimal for the sake of convergence. In this paper, we propose to use multinomial sampling for dropout, i.e.,
Neural Information Processing Systems
Nov-21-2025, 07:17:57 GMT
- Country:
- Europe > Spain
- Andalusia > Granada Province
- Granada (0.04)
- Catalonia > Barcelona Province
- Barcelona (0.04)
- Andalusia > Granada Province
- North America > United States
- Florida > Orange County
- Orlando (0.14)
- Iowa > Johnson County
- Iowa City (0.14)
- Florida > Orange County
- Europe > Spain
- Technology: