Data Science Commons^™

Analyzing Empirical Quality Metrics Of Deep Learning Models For Antimicrobial Resistance, Huy H. Nguyen, Sanjay Pillay, Allison Roderick, Hao Wang, John Santerre

SMU Data Science Review

Antimicrobial Resistance (AMR) is a growing concern in the medical field. Over-prescription of antibiotics as well as bacterial mutations have caused some once lifesaving drugs to become ineffective against bacteria. However, the problem of AMR might be addressed using Machine Learning (ML) thanks to increased availability of genomic data and large computing resources. The Pathosystems Resource Integration Center (PATRIC) has genomic data of various bacterial genera with sample isolates that are either resistant or susceptible to certain antibiotics. Past research has used this database to use ML algorithms to model AMR with successful results, including accuracies over 80%. To better …

Generating And Smoothing Handwriting With Long Short-Term Memory Networks, Muchigi Kimari, Edward Fry, Ikenna Nwaogu, Yumei Bennett, John Santerre

SMU Data Science Review

This project explores the different neural network methods to generate synthetic handwriting text. The goal is to offer an AI tool that generates handwriting, while maintaining an individual’s style, to people suffering with Dysgraphia. As part of this project, an application development framework is setup on GitHub, in such a way that others can continue to explore and improve the AI tool.

Using Deep Learning To Automate The Diagnosis Of Skin Melanoma, Akhil Reddy Alasandagutti

Honors Theses

Machine learning and image processing techniques have been widely implemented in the field of medicine to help accurately diagnose a multitude of medical conditions. The automated diagnosis of skin melanoma is one such instance. However, a majority of the successful machine learning models that have been implemented in the past have used deep learning approaches where only raw image data has been utilized to train machine learning models, such as neural networks. While they have been quite effective at predicting the condition of these lesions, they lack key information about the images, such as clinical data, and features that medical …

Improving Space Efficiency Of Deep Neural Networks, Aliakbar Panahi

Theses and Dissertations

Language models employ a very large number of trainable parameters. Despite being highly overparameterized, these networks often achieve good out-of-sample test performance on the original task and easily fine-tune to related tasks. Recent observations involving, for example, intrinsic dimension of the objective landscape and the lottery ticket hypothesis, indicate that often training actively involves only a small fraction of the parameter space. Thus, a question remains how large a parameter space needs to be in the first place — the evidence from recent work on model compression, parameter sharing, factorized representations, and knowledge distillation increasingly shows that models can be …

Towards High Performance Stock Market Prediction Methods, Warren M. Landis, Sangwhan Cha

Other Student Works

Stock markets of today, and will continue to in the future, rely on the metrics of timeliness and efficiency to reach optimal profits. A way stock investors have continued to strive for the best of these two factors of the business is through the use of predictive machine learning systems to help aid in their decision making. However, among the many systems currently in use, it could be said that the myriad of data that they are based on may not be sufficient. In an effort to devise an ensemble learning predictive system that will utilize an array of big …

Toxic Language Detection Using Robust Filters, Deepti Kunupudi, Shantanu Godbole, Pankaj Kumar, Suhas Pai

SMU Data Science Review

Social networks sometimes become a medium for threats, insults, and other types of cyberbullying. A large number of people are involved in online social networks. Hence, the protection of network users from anti-social behavior is a critical activity [19]. One of the significant tasks of such activity is the detection of toxic language. Abusive/Toxic language in user-generated online content has become an issue of increasing importance in recent years. Most current commercial methods use blacklists and regular expressions; however, these measures fall short when contending with more subtle, lesser-known examples of hate speech, profanity, or swearing[6]. Abusive language classification has …

Multimodal Fusion Strategies For Outcome Prediction In Stroke, Esra Zihni, John D. Kelleher, Vince I. Madai, Ahmed Khalil, Ivana Galinovic, Jochen Fiebach, Michelle Livne, Dietmar Frey

Conference papers

Data driven methods are increasingly being adopted in the medical domain for clinical predictive modeling. Prediction of stroke outcome using machine learning could provide a decision support system for physicians to assist them in patient-oriented diagnosis and treatment. While patient-specific clinical parameters play an important role in outcome prediction, a multimodal fusion approach that integrates neuroimaging with clinical data has the potential to improve accuracy. This paper addresses two research questions: (a) does multimodal fusion aid in the prediction of stroke outcome, and (b) what fusion strategy is more suitable for the task at hand. The baselines for our experimental …

Enhancing Clinical Concept Extraction With Contextual Embeddings, Yuqi Si, Jingqi Wang, Hua Xu, Kirk Roberts

Faculty, Staff and Student Publications

OBJECTIVE: Neural network-based representations ("embeddings") have dramatically advanced natural language processing (NLP) tasks, including clinical NLP tasks such as concept extraction. Recently, however, more advanced embedding methods and representations (eg, ELMo, BERT) have further pushed the state of the art in NLP, yet there are no common best practices for how to integrate these representations into clinical tasks. The purpose of this study, then, is to explore the space of possible options in utilizing these new models for clinical concept extraction, including comparing these to traditional word embedding methods (word2vec, GloVe, fastText).

MATERIALS AND METHODS: Both off-the-shelf, open-domain embeddings and …

Non-Linear Machine Learning With Active Sampling For Mox Drift Compensation, Tamara Matthews, Muhammad Iqbal, Horacio Gonzalez-Velez

Conference papers

Abstract—Metal oxide (MOX) gas detectors based on SnO2 provide low-cost solutions for real-time sensing of complex gas mixtures for indoor ambient monitoring. With high sensitivity under ideal conditions, MOX detectors may have poor longterm response accuracy due to environmental factors (humidity and temperature) along with sensor aging, leading to calibration drifts. Finding a simple and efficient solution to correct such calibration drifts has been the subject of numerous studies but remains an open problem. In this work, we present an efficient approach to MOX calibration using active and transfer sampling techniques coupled with non-linear machine learning algorithms, namely neural networks, …

A Frame-Based Nlp System For Cancer-Related Information Extraction, Yuqi Si, Kirk Roberts

Faculty, Staff and Student Publications

We propose a frame-based natural language processing (NLP) method that extracts cancer-related information from clinical narratives. We focus on three frames: cancer diagnosis, cancer therapeutic procedure, and tumor description. We utilize a deep learning-based approach, bidirectional Long Short-term Memory (LSTM) Conditional Random Field (CRF), which uses both character and word embeddings. The system consists of two constituent sequence classifiers: a frame identification (lexical unit) classifier and a frame element classifier. The classifier achieves an F