Research Frontiers

Thus far, in this book, we have provided an in-depth introduction to multi-modal hash learning for large-scale multimedia retrieval and recommendation. In particular, we first introduce the great demand to develop effective multi-modal hash learning frameworks in big data environments. Then we analyze the prominent research challenges toward this end, such as the heterogeneous modality gap, multi-modal semantic modeling, cold-start and explainable recommendation problem, and ineffective and inefficient hash optimization. To address these issues, we present a series of multi-modal hashing methods, comprising context-aware hashing for image retrieval, cross-modal hashing for unsupervised and supervised cross-modal retrieval, composite multi-modal hashing, and hashing-based multi-modal recommendation methods. Although the above studies have shed some light on large-scale multimedia retrieval and recommendation, we have to admit that this research line is still at the young and up-and-coming stage. Here we list a few promising future research directions with their corresponding challenges.