Visually-Aware Personalized Preference Ranking Network
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional recommendation systems struggle with providing personalized fashion recommendations due to semantic complexity, the cold start problem, and the rapid change in fashion trends, often relying on inefficient data analysis and semantic data that is sparse and unreliable.
Innovation Solution
The implementation of a deep learning-based visually-aware personalized preference ranking network that uses implicit user feedback to train a Siamese convolutional neural network and a personalized ranking model, allowing for the identification of latent item and user features, and providing accurate fashion recommendations by correlating these features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional recommendation systems use semantic data for fashion recommendations, then they can provide personalized recommendations, but the semantic data is sparse and unreliable leading to poor recommendation accuracy
Solution Approach 1:
The patent replaces conventional semantic-based recommendation mechanisms with a deep learning-based visual analysis system. The system uses convolutional neural networks to process fashion item images directly, substituting the mechanical semantic data processing approach with a visual-based deep learning approach that can reliably extract features without depending on sparse semantic annotations.
Solution Approach 2:
The patent introduces visual features as an intermediary between the fashion item and the recommendation system. Instead of directly using sparse semantic data, the system extracts visual features from images that serve as a rich intermediary representation, capturing fashion item characteristics more comprehensively and reliably for personalization.
2Adaptability or versatility
If conventional systems gather and analyze extensive data for personalized recommendations, then they can provide personalized results, but they waste computing resources due to inefficient data analysis
Solution Approach 1:
The patent extracts only the essential visual features needed for fashion recommendation using deep learning models. Instead of gathering and analyzing extensive redundant data, the system extracts relevant visual characteristics directly from images, reducing computational overhead while maintaining personalization capability.
Solution Approach 2:
The patent changes the parameter representation from extensive semantic data to compact visual feature vectors extracted by neural networks. This parameter transformation reduces the dimensionality and complexity of data analysis while preserving the essential information needed for personalized fashion recommendations.
3Productivity
If conventional systems use pre-extracted features for fashion item classification, then they can organize items, but these features are geared toward classification rather than recommendation leading to poor recommendation performance
Solution Approach 1:
The patent implements a dynamic feature extraction system using deep learning models that adapt features specifically for recommendation tasks. Instead of using static pre-extracted classification features, the system dynamically learns and extracts visual features optimized for personalization and recommendation, improving recommendation performance while maintaining organization efficiency.
Data Source
AI summary
The present disclosure relates to a fashion recommendation system that employs a task-guided learning framework to jointly train a visually-aware personalized preference ranking network. In addition, the fashion recommendation system employs implicit feedback and generated user-based triplets to learn variances in the user's fashion preferences for items with which the user has not yet interacted. In particular, the fashion recommendation system uses triplets generated from implicit user data to jointly train a Siamese convolutional neural network and a personalized ranking model, which together produce a user preference predictor that determines personalized fashion recommendations for a user.