- Thursday, November 14, 2019: MLBytes + Freenome: Dr. Joyce Liu
- Ahmadieh Family Grand Hall (Gross Hall 330), 4.30pm-5.30pm
- Despite the public health emphasis on population-level cancer screening in recent decades, adherence remains lower than desired, and cancer is often detected too late for successful treatment. In this talk, we will present Freenome’s multiomics platform that detects key biological signals from blood. The platform integrates assays for cell-free DNA, methylation, and proteins with advanced computational biology and machine learning techniques to understand additive signatures for early cancer detection. Furthermore, we will present biological insights revealed by latent representations from factor analysis and convolutional neural networks of genomic and proteomic data. These methods can be applied to other cancer types to learn biologically meaningful latent representations and to shed light on immune processes involved in cancer biology.
- Thursday, November 7, 2019: MLBytes + Wovenware: Using ML to Prevent Mosquito-Borne Diseases, Leslie De Jesus
- Ahmadieh Family Grand Hall (Gross Hall 330), 4.30pm-5.30pm
- Machine learning seems to be everywhere, across industries and affecting many aspects of daily living. Aside from the practical applications of machine learning in business, new solutions are being put into action to help environmental causes that can improve the lives of people and the planet. Leslie De Jesus will explain how one such solution is being created by Wovenware for the Puerto Rico Science, Technology, & Research Trust, which is working to help prevent the spread of mosquito-borne diseases, such as Zika and Dengue Fever, in Puerto Rico and eventually around the world, by automating the identification and classification of various species of mosquitoes. The end goal is to use AI to streamline the process through which researchers can uncover insights and correlate specific mosquito populations to environmental conditions, helping them identify and isolate mosquitoes that are resistant to FDA-approved insecticides.
- Saturday & Sunday, November 2-3, 2019: Duke Datathon
- Datathon returns. Attack a dataset and come up with an analysis and visualization while racing against the clock. Top teams will win prizes, and all students will have the opportunity to interact with industry and academic sponsors.
- Thursday, October 24, 2019: MLBytes + Wall Street Journal: AI in the Newsroom, Kabir Seth
- Ahmadieh Family Grand Hall (Gross Hall 330), 4.30pm-5.30pm
- The Wall Street Journal newsroom and Dow Jones are heavily invested in AI and machine learning initiatives, including natural language processing and deep learning. The talk will describe how AI is being used in the newsroom for the 2020 elections. Furthermore, Kabir will walk through examples of how he approaches business problems that can be solved with AI and machine learning at the Wall Street Journal.
- Thursday, September 26, 2019: MLBytes + Infinia ML: Document Matching with Neural Networks, John Bralich
- Ahmadieh Family Grand Hall (Gross Hall 330), 4.30pm-5.30pm
- Search engines are used daily across the world. Given a search term, how can relevant results be returned? This talk discusses information retrieval: specifically, the case where the query is a long text document and relevant documents must be selected from a collection of long text documents. Approaches for accomplishing this task will be reviewed, including bag-of-words, TF-IDF, and neural network-based solutions. Practical considerations for implementation will also be discussed.
- Thursday, September 19, 2019: MLBytes + Twitter: A Deep Dive into Twitter's Timeline, Dr. Satanjeev Bano
- Ahmadieh Family Grand Hall (Gross Hall 330), 4.30pm-5.30pm
- Machine learning impacts everything at Twitter, from showing the most relevant content to users to promoting healthier conversations, highlighting trends, and delivering catered advertisements. In this talk, Satanjeev will present one of the largest applications of ML at Twitter: ranking the main feed of Tweets by the relevance of the Tweets to the viewer. He will present the dimensions of the ML problem: the goals, the metrics to optimize for, the features they use, and the network architecture they have developed. The talk will focus on the practical application of ML, and lessons learned delivering ML-based products to millions of users every day.
- Thursday, September 12, 2019: MLBytes + IBM: The Frontiers of Computation, Dr. Mark Ritter
- Ahmadieh Family Grand Hall (Gross Hall 330), 4.30pm-5.30pm
- Classes of computational problems remain difficult or unsolved despite the incredible gains in hardware capability described by Moore's law. With CMOS scaling approaching its apex, what are the potential future technologies that could continue driving performance improvements and even solve some intractable problems? In this talk, Mark will explore both biologically-inspired technology and quantum technology as two new non-Von Neumann computational paradigms that offer promise to expand the frontiers of computation. Machine learning algorithms, including deep learning, are, at their core, based on decades-old ideas inspired by neuroscience. Some previously unsolved problems have been cracked, but even deep learning research seems to be reaching the point of diminishing returns. After a brief overview of the key breakthroughs, Mark will point to some features that, if understood, may yield new inspiration to advance our algorithms toward more powerful ones.
- Thursday, April 18, 2019: MLBytes: Analyzing Human Rights Abuses using Bayesian Entity Resolution, Neil Marchant
- Old Chem 116, 11.45am-1pm
- Accurate statistics on human rights abuses are essential for supporting evidence-based policy, measuring progress, and raising awareness in the general public. In this talk, Neil Marchant will discuss unsupervised Bayesian entity resolution models, which are able to identify duplicate records in the data, while quantifying uncertainty. He will highlight the importance of choosing flexible priors and in implementing scalable inference algorithms, and will present preliminary results on synthetic data and a data set containing death records from the Salvadoran Civil War. Neil is a third-year Ph.D. candidate in the School of Computing and Information Systems at the University of Melbourne, Australia.
- Thursday, April 11, 2019: MLBytes + Microsoft: ML and Financial Forecasting, Amita Gajewar
- Ahmadieh Family Grand Hall (Gross Hall 330), 4.30pm-5.30pm
- Multiple finance organizations within Microsoft rely on having accurate financial forecasts for critical business decisions. One such organization, the Enterprise Partner Group, serves some of the largest commercial and public sector organizations across the globe, with total product revenue in the range of billions of dollars worldwide. Generating up-to-date, accurate predictions with manual forecast methods can be challenging; Microsoft Finance sought a solution that can rely on state-of-the-art machine learning algorithms. In this talk, Amita Gajewar will explain revenue forecasting models that she developed to forecast quarterly revenue for enterprise products in several regions across the globe and their world-wide aggregate. Amita is a senior data scientist in the Cloud + AI group at Microsoft. Previously, she has worked on various projects in computational advertising at Microsoft and Yahoo!. Amita graduated from the Georgia Institute of Technology with a Masters in Computer Science.
- Thursday, March 28, 2019: MLBytes + Microsoft Research: Artificial Emotional Intelligence, Dr. Daniel McDuff
- Ahmadieh Family Grand Hall (Gross Hall 330), 4.30pm-5.30pm
- Emotions play an important role in our everyday lives. They influence memory, decision-making, and well-being. In order to advance the fundamental understanding of human emotions, build smarter affective technology, and ultimately help people, we need to perform research in-situ. Leveraging exciting advances in machine learning, it is now possible to quantify emotional responses on a large scale using webcams and microphones in everyday environments. This talk describes novel methods for physiological and behavioral measurement via ubiquitous hardware, including state-of-the-art approaches for emotion synthesis that can be used to create rich human-agent or robot interactions. Dr. Daniel McDuff is a researcher at Microsoft Research, where he leads research and development of affective computing technology, with a focus on scalable tools to enable the automated recognition and analysis of emotions and physiology. Daniel completed his Ph.D. in the Affective Computing Group at the MIT Media Lab in 2014 and has a B.A. and Masters from Cambridge University.
- Saturday, March 23, 2019: Duke Machine Learning Day
- Come to Duke's second Machine Learning Day on Saturday, March 23 in Schiciano Auditorium and The Edge! At ML Day, you'll have the opportunity to learn about exciting machine learning research through workshops, talks, and research presentations on machine learning and related topics. This year's event will highlight important women in the field, including a women in data science dinner.
- Thursday, March 21, 2019: MLBytes: Classical Music Composition Using State Space Models, Anna Yanchenko
- Ahmadieh Family Grand Hall (Gross Hall 330), 4.30pm-5.30pm
- Algorithmic composition of music has a long history, and with the development of powerful deep learning methods, there has recently been increased interest in exploring algorithms and models to create art. We explore the utility of state space models, in particular, hidden Markov models, in composing classical piano pieces. Anna Yanchenko is a Ph.D. student in the Department of Statistical Science at Duke. She is interested in machine learning for time series applications, especially with applications to modeling music.
- Wednesday, February 27, 2019: Citadel Quantitative Trading Challenge
- Schiciano Auditorium, 6.30pm-8.30pm
- Learn about opportunities in investment and trading, quantitative research, and software engineering at Citadel and Citadel Securities, and participate in a quantitative trading challenge, co-hosted by Duke Undergraduate Machine Learning. More information can be found here.
- Thursday, February 21, 2019: MLBytes + Infinia ML: Word Embeddings for Neural Network based NLP, John Bralich
- Ahmadieh Family Grand Hall (Gross Hall 330), 4.30pm-5.30pm
- Neural network based approaches have achieved state-of-the-art results on many benchmark natural language processing tasks. These approaches rely on the representation of words as vectors, also called word embeddings. In this talk, different approaches for the creation of word embeddings will be discussed, as well as popular models for generating these embeddings like Word2Vec, ELMo, and others. John Bralich is currently a data scientist at Infinia ML. Prior to this, he worked for a Durham based startup as a machine learning engineer. There he applied machine learning techniques to perform medical document analysis with the goal of creating more patient-friendly medical reports. He received his Bachelor's and Master's in Electrical and Computer Engineering from Duke University.
- Thursday, February 7, 2019: MLBytes + Tresata: Fundamentals of Distributed Machine Learning, Koert Kuipers
- Ahmadieh Family Grand Hall (Gross Hall 330), 4.30pm-5.30pm
- Koert is the Chief Technology Officer of Tresata, an analytics software company and unicorn startup. Within Tresata, he leads the development of the core distributed data processing, analytics, and machine learning platform. This talk will explore the building blocks of distributed machine learning, including modern distributed platforms, the transformations they provide, and how they can be used to efficiently implement highly scalable machine learning algorithms. This is useful for data scientists that do not plan to implement algorithms themselves, as it gives them a sense of the performance of algorithms on these platforms.
- Thursday, November 15, 2018: MLBytes: Classification and Topology, Olivier Binette
- Ahmadieh Family Grand Hall (Gross Hall 330), 4.30pm-5.30pm
- Mathematicians of the 20th century developed topology as a tool to study qualitative properties of shapes and surfaces. This talk will introduce some of their ideas and show how they can be used to gain useful insight in binary classification. Following classical theorems, we will obtain a "topologically accurate" reconstruction of class boundaries and hint towards current research. Olivier Binette is a M.Sc. student at Université du Québec à Montréal.
- Friday, November 2, 2018: MLBytes: Addressing Human Trafficking with Machine Learning, Matthew Barnes
- Ahmadieh Family Grand Hall (Gross Hall 330), 12pm-1pm
- Over the past decade, Carnegie Mellon University has scraped and indexed billions of online escort advertisements for the purpose of policing human trafficking. The huge volume and complexity of data makes it impossible for law enforcement officers to manually digest. In this talk, Matt will reveal how he used a suite of machine learning tools, including deep learning, natural language processing, record linkage, and massively scalable hashing techniques, to provide human trafficking advertisement search, analysis, and prediction capabilities to dozens of law enforcement agencies across the country. Matt received his M.S. in Robotics and is currently pursuing his Ph.D. at Carnegie Mellon University, where he studies foundational machine learning, including clustering and detecting bias in groups of data.
- Saturday, October 27, 2018: Duke Datathon
- Duke's datathon. Attack a dataset and come up with an analysis and visualization while racing against the clock. Top teams will win prizes, and all students will have the opportunity to interact with industry and academic sponsors.
- Thursday, October 25, 2018: MLBytes + Infinia ML: Real-time Object Detection, Dr. Ikenna Odinaka
- Ahmadieh Family Grand Hall (Gross Hall 330), 4.30pm-5.30pm
- YOLO (You Only Look Once) is a real-time object detection system that can be used to determine the positions of over 9000 types of objects in images. In this exciting computer vision talk by Dr. Ikenna Odinaka, we'll explore the architecture and techniques behind the state-of-the-art YOLO algorithm. Dr. Odinaka is a data scientist at Infinia ML, and previously studied Electrical Engineering at Duke University and Washington University in St. Louis.
- Thursday, October 18, 2018: MLBytes + Infinia ML: Visual Dialog, Dr. Justin Ellis
- Ahmadieh Family Grand Hall (Gross Hall 330), 4.30pm-5.30pm
- Visual dialog is an exciting new field in AI that requires an AI agent to hold meaningful dialog with humans in natural, conversational language about visual content, combining techniques from computer vision, natural language processing, and dialog systems research. Dr. Ellis is a data scientist at Infinia ML. Before joining Infinia ML, Justin was an Einstein Postdoctoral Fellow at the NASA Jet Propulsion Laboratory and California Institute of Technology. His research focused on analysis of astrophysical data for gravitational wave detection, and he led a multi-institute team of researchers in statistical analysis and software development.
- Thursday, September 20, 2018: MLBytes: March MATHness, Dr. Tim Chartier
- Ahmadieh Family Grand Hall (Gross Hall 330), 4.30pm-5.30pm
- Every March, there is a lot of madness around "Who's number one?" The Division 1 NCAA men's basketball tournament, often called March Madness, begins! Millions of brackets are created in an attempt to predict the outcomes of the tournament. Who will win round by round? In this talk, we will see how research in ranking algorithms created brackets for March Madness that beat over 90% of over 8 million brackets submitted to ESPN's online tournament. Have a sport you'd like to analyze? By the end, you'll know how to create your own ranking method to answer "Who's number one?"
- Fall 2018: Applied Machine Learning
- Take a half-credit course at Duke in machine learning, taught by co-president Shrey Gupta. The course includes learning specific machine learning concepts (classification, regression, clustering, object detection, dimensionality reduction, and evaluation of models), their historical origins, and existing and potential applications, and will be taught during the Fall 2018 semester. Register on DukeHub.
- Saturday, March 31, 2018: Duke Machine Learning Day
- Come to Duke's first-ever Machine Learning Day, for undergraduates, on Saturday, March 31 in Gross Hall! At ML Day, you'll have the opportunity to learn about Duke's exciting machine learning research. And, if you're a student researcher in ML, you'll have the opportunity to present your work!
- Monday, March 19, 2018: Data Science at Facebook
- Old Chem 116, 11.45am-12.45pm
- Learn about data science at Facebook, co-hosted by Duke Statistical Science and Duke Undergraduate Machine Learning.