Use of Learned Data Structures in Machine Learning and AI Algorithms
Author : Rishik Jariwala and Radhika Patwardhan
Abstract :
The rapid growth of modern data has driven a shift from static storage designs to intelligent, data-aware architectures. Traditional storage systems struggle with the sparse and high-dimensional data used in AI, creating the need for advanced solutions such as multidimensional indexing, sparse tensors, and graph-based frameworks. This evolution introduced learned data structures, which apply machine learning to predict data locations by modeling underlying data distributions. Predictive indexes like the PGM-index and Recursive Model Index replace rigid tree structures, significantly improving memory efficiency and search performance.[2] However, these advances also demand changes in data science education. Academic curricula are moving away from outdated procedural languages toward flexible platforms like Python, enriched with real-world case studies from distributed systems and search technologies. Hands-on, multi-level experimental environments prepare students to manage large-scale data effectively. Integrating predictive storage architectures with modern teaching approaches equips future professionals to handle today’s complex and data-intensive information landscape efficiently.
Keywords :
Data-aware architectures, learned data structures, machine-learning indexing, PGM-index, Recursive Model Index, sparse high-dimensional data.
