Cookie-Space Domain Adaptation for Cookieless Device Attributes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine-learning models trained on client devices that allow record keeping or analysis produce inconsistent results when classifying devices that block such techniques, leading to inaccurate content selection and provision on devices that prevent cookie-based data collection.
Innovation Solution
Utilize domain adaptation techniques like importance reweighting, maximum mean discrepancy, correlational alignment, and reverse gradient to train models on known client devices with cookies to predict attributes of devices that block cookie analysis, ensuring consistent performance across different cookie spaces.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a machine-learning model is trained on client devices that allow record keeping or analysis, then the model can accurately classify devices with cookies, but the model produces inconsistent results when classifying devices that block cookie analysis
Solution Approach 1:
The patent applies domain adaptation techniques that modify model parameters and training objectives to account for differences between cookie-based and cookieless environments. By changing the training parameters to include domain adaptation loss functions and reweighting schemes, the model achieves consistent performance across both device types without sacrificing accuracy on either domain
Solution Approach 2:
The patent introduces an intermediary domain adaptation layer that mediates between the source domain (cookie-based devices) and target domain (cookieless devices). This intermediary mechanism transfers knowledge from the source domain to the target domain through techniques like feature alignment and gradient reversal, enabling the model to generalize across domains while maintaining reliability
2Ease of manufacture
If a model is trained to classify devices with known attributes, then training data is readily available, but the model cannot accurately predict attributes of devices that block cookie analysis
Solution Approach 1:
The patent performs preliminary domain adaptation training using available cookie-based device data before deploying the model to cookieless environments. By pre-training on the source domain and then adapting to the target domain through techniques like importance reweighting and maximum mean discrepancy, the model prepares in advance for the cookieless scenario, enabling accurate attribute prediction even without direct training data from blocked devices
Solution Approach 2:
The patent modifies training parameters to simulate cookieless device characteristics during the adaptation phase. By changing optimization parameters and introducing domain confusion techniques, the model learns to generalize from cookie-based data to predict attributes of cookieless devices accurately, bridging the data availability gap
3Adaptability or versatility
If content selection is based on cookie-based device classification, then personalization is improved for devices with cookies, but content delivery is inaccurate for devices that block cookie analysis
Solution Approach 1:
The patent creates a universal content selection model that functions accurately across both cookie-based and cookieless devices. By implementing domain adaptation throughout the content delivery pipeline, the single model achieves multi-functionality, providing personalized content selection for devices with cookies while maintaining accurate content delivery for devices that block cookie analysis, without requiring separate models for each device type
Data Source
AI summary
The present disclosure describes techniques for training a model using cross-domain adaptation to classify content requests from client devices having unknown attributes. The system can obtain requests for content from client devices of a first domain, and requests for content from client devices of a second domain. The system can train a model by propagating request attributes of the first domain through the model to generate first internal data from an internal layer of the model and a first output vector an output layer of the model. The system can propagate request attributes of the second domain to generate second internal data, and determine a difference between the first internal data and the second internal data. The system can update the model based on the difference and the output vector, and classify a third client device of the second domain using the model.