Andrew Jacobsen

Papers

Total Citations

H-Index

About

Andrew Jacobsen is a researcher specializing in online learning, continual prediction, and adaptive optimization algorithms. His work focuses on improving the stability and efficiency of machine learning models in non-stationary environments, where data distributions shift over time. Jacobsen’s most cited paper, "Meta-Descent for Online, Continual Prediction" (2019), investigates vector step-size adaptation methods to enhance stochastic gradient descent in continual learning settings. By demonstrating how adaptive step-size scaling can significantly outperform vanilla SGD, his research provides practical solutions for real-time prediction tasks that must evolve with streaming data. With 11 citations, this work has influenced subsequent studies in meta-learning and adaptive optimization. Jacobsen’s contributions are particularly relevant for applications in robotics, finance, and autonomous systems, where models must continuously learn from changing inputs without catastrophic forgetting. His research bridges theoretical insights with algorithmic innovations, offering robust tools for lifelong learning systems.