Information Processing Apparatus for Cold Start Recommendation via Similarity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional recommendation systems face challenges in reducing the impact of the cold start problem, particularly due to insufficient history information, and existing techniques do not adequately consider user viewpoints when transferring knowledge between models.
Innovation Solution
An information processing apparatus that acquires content-to-content and user-to-user similarities to estimate prior distributions of expected rewards, using a bandit algorithm to derive posterior distributions and determine suitable content for users, thereby addressing the cold start issue by transferring knowledge across domains and users based on recent actions and similarities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If collaborative filtering is used to determine similarity between users based on item transaction history, then recommendation effect is improved, but the cold start problem occurs when history information is insufficient
Solution Approach 1:
The patent applies preliminary action by pre-training a teacher model on abundant data from a source domain before deploying it to the target domain. This pre-trained model serves as a knowledge base that can provide preliminary recommendations even when target domain history information is insufficient, thereby addressing the cold start problem while maintaining recommendation quality
Solution Approach 2:
The patent introduces a teacher-student model framework where the teacher model acts as an intermediary between the source domain knowledge and the target domain application. The teacher model transfers learned patterns and representations to the student model, enabling the system to overcome the lack of sufficient history information in the target domain while maintaining reliable recommendations
2Adaptability or versatility
If knowledge transfer is performed between independent domain-specific models, then cross-domain recommendation capability is improved, but user viewpoints are not adequately considered
Solution Approach 1:
The patent implements universality by designing a unified teacher-student framework that can handle multiple domains and user perspectives simultaneously. The teacher model learns universal patterns from the source domain that can be adapted to various target domains, while the framework preserves user-specific information through careful knowledge transfer mechanisms that consider individual user viewpoints rather than treating all users uniformly
Solution Approach 2:
The patent applies local quality by selectively transferring knowledge at different levels: global domain knowledge is transferred from the teacher model, while local user-specific information is preserved and adapted in the student model. This allows the system to maintain cross-domain adaptability while preserving important user viewpoint information that would otherwise be lost in generic knowledge transfer
Data Source
AI summary
An information processing apparatus (100) includes: an acquisition unit configured to acquire a content-to-content similarity that is a similarity between a target content and one or more pieces of other content, and a user-to-user similarity that is a similarity between a target user and one or more other users; an estimation unit configured to estimate a prior distribution of expected rewards obtained as a result of execution processing performed by the target user on the target content, based on the content-to-content similarity and the user-to-user similarity; and a derivation unit configured to derive a posterior distribution of the expected rewards, using the prior distribution.