Narrowing down millions of videos (like your .mp4 file) to a few hundred based on user history.
While that specific filename does not appear in standard academic databases, the term in this context typically refers to one of two things: _-4969924009768.mp4
Knowing if it came from a specific GitHub repo, a dataset, or a social media metadata leak would help me identify the exact research paper you're looking for. Narrowing down millions of videos (like your
Researchers often track these on the Deep Learning Paper Reading Table on GitHub. Weighting training examples by watch time rather than
1. The "Deep Neural Networks for YouTube Recommendations" Paper
If the filename is from a specific dataset (like Kinetics-700 or UCF101), "deep paper" likely refers to research on or Action Recognition . These papers focus on using neural networks to "understand" or compress video files efficiently.
Weighting training examples by watch time rather than just clicks, which revolutionized how "viral" or relevant videos are identified. You can read the full text at Google Research . 2. Deep Video Compression/Analysis
