Enhancing the Cognition and Efficacy of Machine Learning Through Similarity

• Published in: SN computer science Vol. 3; no. 6; p. 442
• Main Authors: Pendyala, Vishnu, Amireddy, Rakesh
• Format: Journal Article
• Language: English
• Published: Singapore Springer Nature Singapore 01.11.2022; Springer Nature B.V
• Subjects: Algorithms; Brain; Cognition; Computer Imaging; Computer Science; Computer Systems Organization and Communication Networks; Data Structures and Information Theory; Deep learning; Information Systems and Communication Service; Machine learning; Neural networks; Original Research; Pattern Recognition and Graphics; Physics; Programming languages; Python; Similarity; Soft Computing for Real Time Engineering Applications; Software Engineering/Programming and Operating Systems; Support vector machines; Teaching; Teaching machines; Vision; Deep learning; Similarity; Machine learning; Transcoding
• ISSN: 2661-8907; 2662-995X; 2661-8907
• DOI: 10.1007/s42979-022-01339-y

Similarity is a key element of machine learning and can make human learning much more effective as well. One of the goals of this paper is to expound on this aspect. We identify real-world concepts similar to hard-to-understand theories to enhance the learning experience and comprehension of a machine learning student. The second goal is to enhance the work in the current literature that uses similarity for transcoding. We uniquely try transcoding from Python to R and vice versa, something that was not attempted before, by identifying similarities in a latent embedding space. We list several real-world analogies to show similarities with and simplify the machine learning narrative. Next, we use Cross-Lingual Model Pretraining, Denoising Auto-encoding, and Back-translation to automatically identify similarities between the programming languages, Python and R and convert code in one to another. In the course of teaching machine learning to undergraduate, graduate, and general pool of students, the first author found that relating the concepts to real-world examples listed in this paper greatly enhanced student comprehension and made the topics much more approachable despite the math and the methods involved. When it comes to transcoding, in spite of the fact that Python and R are substantially different, we obtained reasonable success measured using various evaluation metrics and methods as described in the paper. Machines and human beings predominantly learn by way of similarity, a finding that can be explored further in both the machine and human learning domains.