Iterative relevance feedback with adaptive exploration/exploitation trade-off

Abstract

Content-based image retrieval systems have to cope with two different regimes: understanding broadly the categories of interest to the user, and refining the search in this or these categories to converge to specific images among them. Here, in contrast with other types of retrieval systems, these two regimes are of great importance since the search initialization is hardly optimal (i.e. the page-zero problem) and the relevance feedback must tolerate the semantic gap of the image's visual features.We present a new approach that encompasses these two regimes, and infers from the user actions a seamless transition between them. Starting from a query-free approach meant to solve the page-zero problem, we propose an adaptive exploration/exploitation trade-off that transforms the original framework into a versatile retrieval framework with full searching capabilities. Our approach is compared to the state-of-the-art it extends by conducting user evaluations on a collection of 60,000 images from the ImageNet database.

A Pseudo Feedback-Based Annotated TF-IDF Technique for Dynamic Crypto-Ransomware Pre-Encryption Boundary Delineation and Features Extraction

“…At the outset, an exploration process was carried out to build an initial boundary vector called out of the explicit CryptoAPI calls, i.e. the cryptography-specific APIs, according to [40]. These cryptoAPIs were used as seeds for the preencryption boundary vector.…”

Query-Free Clothing Retrieval via Implicit Relevance Feedback

“…Afterwards, Ferecatu [16] extended the framework to category search instead of target search. The application was scaled to large-scale datasets by Suditu and Fleuret [47], [48] who adopted a hierarchical and expandable adaptive trace algorithm benefited from adaptive exploration/exploitation tradeoff. Similar to the idea of mental image retrieval, Auer et al [3] maintained the weights of images by giving less relevant images a constant discount at each iteration.…”

“…Several recent studies have attempted to scale up mental image retrieval applications [16], [47], [48]. In fact, prior to [15], such algorithms were mainly extended by applying predefined clustering for categorical searches and by applying a tree-and-trace approach for Voronoi partitioning, both of which are complementary to our algorithm.…”

“…the absence of an explicit query image, we propose to solve this problem via implicit relevance feedback [58]. Our method is inspired by the mental image retrieval algorithm [15], [16], [47], [48], which introduced an interactive search paradigm based on user feedback. In this method, the target image is modeled as a random variable, and a cumulative probability distribution of this variable is stored to track past interactions with the user.…”

Exquisitor

“…Through feedback from the user, interactive learning can adapt to the intent and knowledge of the user, and thus collaborate with the user towards, e.g., learning new or unknown analytic categories on the fly. Previous contributions, however, largely operated at a relatively small scale [1,2,8,10,13,14].…”