Anirudh Karnam
Pronouns: he/him
Research Mentor(s): Cristian Minoccheri
Research Mentor School/College/Department: Computational Medicine and Bioinformatics / Medicine
Program: UROP
Session: Session 5 (2:40pm – 3:30pm)
Authors: Anirudh Karnam, Cristian Minoccheri
Abstract
Neural networks require training data to learn and improve their accuracy over time, but once these algorithms are fine-tuned, they are powerful tools in artificial intelligence. The purpose of this research is to test a novel synergy between neural networks and tensor methods, by comparing neural networks for the classification of the MNIST dataset: one fully neural connected neural network, one fully connected neural network with tensor compression of the layers, one convolutional neural network, and one convolutional neural network with a tensor regression layer. This is expanding on the knowledge of coding in Python and linear algebra applications. The ideas we are using in this research use machine learning techniques and tensor methods which can be used in other applications as well. Neural networks can help computers make intelligent decisions with limited human assistance. This is because they can learn and model the relationships between input and output data that are nonlinear and complex. Neural networks, which are a form of artificial intelligence, can be applied to image classification and are currently the state of the art methodology in this field. In the biomedical domain, it is important to analyze images to extract complex features, allowing for better and more effective treatment of patients and differentiating them from patients who may have other disorders or those who are healthy. The results of this research focus on the effects of combining tensor methods with neural networks on the benchmark MNIST dataset.
