ARTFEED — Contemporary Art Intelligence

Joint Spatio-Temporal Enlargement for Micro-Video Popularity Prediction

digital · 2026-04-24

A recent study published on arXiv (2604.20311) introduces a comprehensive framework for predicting the popularity of micro-videos (MVPP), enhancing both temporal and spatial dimensions simultaneously. Current MVPP techniques face challenges due to sparse short-range temporal sampling and a limited capacity for retrieval memory. The proposed framework allows for accurate analysis of very long video sequences and features a scalable memory bank capable of infinite expansion to include all pertinent historical information. This advancement seeks to enhance content recommendations and optimize traffic distribution on digital media platforms.

Key facts

  • Paper ID: arXiv:2604.20311
  • Announce Type: cross
  • Focus: micro-video popularity prediction (MVPP)
  • Goal: forecast future popularity of videos on online media
  • Applications: content recommendation, traffic allocation
  • Limitations addressed: sparse short-range temporal sampling, flat retrieval memory with limited capacity
  • Proposed solution: joint spatio-temporal enlargement framework
  • Enables precise perception of extremely long video sequences
  • Supports infinitely expandable scalable memory bank

Entities

Institutions

  • arXiv

Sources