FURI | Fall 2018

Pattern Extrapolation of Integer Sequences using Novel Machine Learning Techniques

Education icon, disabled. A purple mortarboard.

This project focuses on a deceptively simple problem in machine learning: The extrapolation and learning of integer sequences by a computer program. (For example, learning and extending the fibonacci numbers). While this problem is simple, explicit modeling and pattern extrapolation are at the heart of many interesting problems (like Raven’s progressive matrices, which is an IQ test item). Pattern-extrapolation doesn’t yet have a satisfactory solution, and it is an open question if traditional machine learning approaches are an efficient approach for this problem. This project compares advanced approaches to the problem, and proposes a more promising algorithm.

Student researcher

Lucas Saldyt

Computer science

Hometown: Mesa, Arizona

Graduation date: Spring 2020