The ACM Southeast (ACMSE) conference is the oldest, continuously running, annual conference of the Association for Computing Machinery (ACM). ACMSE provides an excellent forum for students, faculty, researchers, and industry practitioners to present their research in a friendly and dynamic atmosphere in all areas of computer science.

ACMSE 2021 is organized by the Mathematical, Computing, & Information Sciences Department (MCIS Department) at Jacksonville State University (JSU), Jacksonville, Alabama. It was held virtually this year on April 15-17, 2021.

Vaidyanath Areyur Shanthakumar, Computer Science PhD candidate, presented during Session 5: Neural Networks, Smartphones, and Games and was awarded Best Full Paper at this year's conference:

Item Based Recommendation Using Matrix-factorization-like Embeddings from Deep Networks
Vaidyanath Shanthakumar, Clark Barnett, Keith Warnick, Putu Sudyanti, Vitalii Gerbuz, Tathagata Mukherjee 

In this paper we describe a method for computing item based rec- ommendations using matrix-factorization-like embeddings of the items computed using a neural network. Matrix factorizations (MF) compute near optimal item embeddings by minimizing a loss that measures the discrepancy between the predicted and known values of a sparse user-item rating matrix. Though useful for recommenda- tion tasks, they are computationally intensive and hard to compute for large sets of users and items. Hence there is need to compute MF-like embeddings using other less computationally intensive methods, which can be substituted for the actual ones. In this work we explore the possibility of doing the same using a deep neural net- work (DNN). Our network is trained to learn matrix-factorization- like embeddings from easy to compute natural language processing (NLP) based semantic embeddings. The resulting MF-like embed- dings are used to compute recommendations using an anonymized user product engagement dataset from the online retail company We present the results of using our embeddings for computing recommendations with the produc- tion dataset consisting of ∼3.5 million items and ∼6 million users. Recommendations from’s own recommendation system is compared against those obtained by using our MF-like embeddings, by comparing the results from both to the ground truth, which in our case is actual user co-clicks data. Our results show that it is possible to use DNNs for efficiently computing MF- like embeddings which can then be used in conjunction with the NLP based embeddings to improve the recommendations obtained from the NLP based embeddings.

The full paper will be published in the ACM Digital Library.