A Transfer Learning Load Adjusted Approach for Video-on-Demand Systems Given Limited Training Data
Year: 2025
Type: conference-paper
Abstract: Inadequate data complicates planning and allocation of VoD resources, potentially hindering the scalability of VoD services. We propose a Transfer Learning Load Adjusted (TLLA) algorithm for resource management given limited VoD data. TLLA leverages the knowledge gained from pre-trained models by storing features and patterns that can be used to train Machine Learning (ML) related tasks. We model limitations in VoD data by proportionally freezing 50% of the neural layers in models trained from pre-trained and source domains. We evaluate the performance of the frozen neural layers by comparing them to unfrozen data. Freezing 50% of the neural layers in pre-trained model shows ≈ 80% similarity to model frozen from source data. These findings have the potential to contribute knowledge that could be utilized when inadequacies in VoD resources are encountered.
DOI: 10.21427/szb5-t470
Authors: Kimeli, Kangogo, de Frein, Ruairi
Institution: Technological University Dublin
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7081208530
FWCI:
Cites:
Cited by:
Related to: