• Powering the Ultimate Smart Driving Machine with Machine Learning

    Abstract: Cars are now connected to the internet and car manufacturers are becoming more like software companies, shifting focus from traditional hardware platforms towards the software and services platform, leveraging scalable cloud computing with machine learning and IoT. The car is not just about going from point A to point B anymore, but rather delivering a holistic experience that revolves around the activities in our daily lives. What are the challenges and opportunities that big data and machine learning can address? Learn what BMW is doing with machine learning to address the challenges of journey management, integrating daily activities into the car, and performing preventative automotive maintenance, to make the car become The Ultimate Smart Driving Machine. Bio: Dr. Alvin Chin is a Senior Researcher in Machine Learning at BMW Technology Corporation, Chicago. His research interests include connected car, machine learning, big data, mobile social networking, and ubiquitous computing. Dr. Chin has authored more than 30 publications and 10 patents, including pending. He received a PhD in Computer Science from the University of Toronto and previously worked for Nokia Research Center, Nokia and Microsoft in Beijing, China. Dr. Chin is a member of various program committees such as IEEE VTC and ACM KDD. Dr. Chin is also active in the Chicago community, as Vice-Chair of ACM Chicago and President of the IEEE Vehicular Technology Society Chicago Chapter. He is the Industry Sessions Co-Chair of the IEEE Vehicular Technology Conference 2019-Fall and was Co-Chair of the Industry Track for IEEE Vehicular Technology Conference-Fall 2018. ______________________________________________________________________________________ 6:00 p.m - 6:30 p.m is time for social. Seminar will start at 6:30 p.m. Please note the office security needs to check your government issued ID (Driver License or Passport, please note Student ID is not valid for entry) . Thanks for your understanding. Our Sponsor: Allstate ( http://www.allstatedatascience.com/ ) Please note we move the Meetup from Suite 875 to Suite 850 (same floor, but on the Allstate-branded side).

  • PyData May Meetup - Process Mining

    Allstate

    Abstract: Process Mining describes a set of data-driven techniques to discover process models from event logs, to check conformance and monitor processes, and to enhance and improve models. During the last ten years, these techniques have developed into a mature research field with many use cases, resulting in a strong industry uptake. Its main applications can be found in business process management to create actionable insights leveraging existing data from IT systems. However, recent research shows a broader range of Process Mining applications such as in Software Engineering, Human-Computer Interaction, and control of manufacturing processes. In this presentation, Julian Theis will introduce the fundamentals of Process Mining based on real-world examples. He will also talk about recent research projects in atypical fields and will discuss future directions. Bio: Julian Theis is currently pursuing the Ph.D. degree with the Mechanical and Industrial Engineering Department, University of Illinois at Chicago (UIC). He received the B.Eng. degree in Media Technology in 2014 and the M.Eng. degree with distinction in Media and Communications Technology in 2016, both from the RheinMain University of Applied Sciences, Wiesbaden, Germany. Julian joined PROMINENT Laboratory, UIC's foremost research facility in Process Mining, as a researcher in January 2017, after working as a software engineer for a leading company in media and workflow automation, broadcast management, and OSS software in Germany. His research interests include Process Mining, its applications to industrial, IT, and business processes, and Deep Learning. Julian is furthermore an IEEE Graduate Student Member and received the IABM Engineering Student Award in 2013. ______________________________________________________________________________________ 6:00 p.m - 6:30 p.m is time for social. Seminar will start at 6:30 p.m. Please note the office security needs to check your government issued ID (Driver License or Passport, please note Student ID is not valid for entry) . Thanks for your understanding. Our Sponsor: Allstate ( http://www.allstatedatascience.com/ ) Please note we move the Meetup from Suite 875 to Suite 850 (same floor, but on the Allstate-branded side).

    1
  • PyData March Meetup - Two Cases Show

    Allstate

    In PyData March event, we are pleased to invite three speakers Doug MacLeod, Heather Wigell, and Itay Livni to demonstrate two separate data/software solutions under business/education domains. Schedule: 1. Between 6:20 p.m. and 7:00 p.m. Topic: What constitutes “useful” or “worth it”. Using visualizations to communicate possibilities of data science to sales executives. Abstract: SCF is creating a predictive order technology for CPG wholesalers. It predicts what products will be ordered by a salesperson for beverage wholesalers in a large format grocery on a given day. It estimates the number of cases a salesperson will order (product quantity), each time a salesperson puts in an order. We are working with an obfuscated transactional data set from a major CPG beverage wholesaler. Join us to discuss how and why we made our choices, the issues we ran into, and how we are solving problems by using Tableau data visualizations to communicate to the business. A discussion of our data and pipeline will be included. See data visualizations of actual versus predicted orders. Bio Doug MacLeod has spent over 20 checkered years in IT, ranging from applications development consultant to Director of Applications Development at Pepsi. He has spent time at major beverage manufacturing and distribution companies such as Pepsi, DS Waters and other distributors specializing in mobile field workforce computing, warehouse and delivery operations. Technical expertise includes all major databases, Objective C, some Java and now just another newbie in Python. Heather Wigell is the Co-Founder at Smart Cloud Forge, and designs data visualizations on Tableau. Heather has worked in marketing for 10 years. Through out her career she has taken on roles in research, analytics, data science and data visualization. Past projects include The Digital Publics Project at University of Chicago. 2. Between 7:00 pm and 8:00 pm Topic: Building an Automated Educational Vocabulary Game Generator Abstract: Vocabulary, repetition, and examples are fundamental to human learning. These fundamental tools help humans to teach each other, communicate, and innovate. Yet, vocabulary building and reading comprehension games specifically geared for science, technology, engineering and math (STEM) disciplines are lacking. They are expensive to make and the content can have a short life span. In this talk we will go over the main components required to build an automated educational vocabulary game generator with packages such as pandas, textacy, gensim, scikit-learn, and networkx. Bio: Itay Livni currently is working on tapNotion full time. A system to automatically generate educational STEM vocabulary games. In his past, he traded equity options in trading pits with hand signals as a market maker and specialist, ran a small proprietary black box trading firm, and worked as a clerk on the CME. With mixed success he tried to automate every job he did. ______________________________________________________________________________________ 6:00 p.m - 6:20 p.m is time for social. Seminar will start at 6:20 p.m. Please note the office security needs to check your government issued ID (Driver License or Passport, please note Student ID is not valid for entry) . Thanks for your understanding. Our Sponsor: Allstate ( http://www.allstatedatascience.com/ ) Please note we move the Meetup from Suite 875 to Suite 850 (same floor, but on the Allstate-branded side).

    9
  • A birds eye view of Recurrent Neural Networks

    Abstract: In this tutorial talk I will introduce Recurrent Neural Networks (RNNs) - used to perform machine learning tasks on temporal data - with an emphasis on explaining the fundamental dynamic models that underly them. A plethora of examples will be given - complete with "soup to nuts" implementations in Python. Bio: Jeremy Watt runs a local ML consultancy called Degree Six and is a Professor at Northwestern where he teaches a range of courses in machine learning, deep learning, mathematical optimization, and reinforcement learning / automatic control. He has just returned to Earth after an intense 18 month stint writing the second edition of his textbook - Machine Learning Refined (published by Cambridge University Press) - which will be released mid-2019. Schedule: 6:00 p.m - 6:30 p.m is time for social. Seminar will start at 6:30 p.m. Please note the office security needs to check your government issued ID (Driver License or Passport, please note Student ID is not valid for entry) . Thanks for your understanding. Our Sponsor: Allstate ( http://www.allstatedatascience.com/ ) Please note we move the Meetup from Suite 875 to Suite 850 (same floor, but on the Allstate-branded side).

    10
  • Beer, Baseball, Bayes.

    Arity

    Abstract: In this presentation, we'll be study the basics of Bayesian processes, hierarchical models, and analysis using Stan, a probabilistic programming language written in C++. As examples, we will compare the growth of beer markets in the United States and try to plan pitch location based on our knowledge of how pitchers throw and batters react. Bio: Nicholas Ver Halen is a sports analytics enthusiast who is searching for as many statistical tools as possible to better understand and model the decision-making process of athletes and coaches. He developed PytchFX, a python implementation of the PitchRx, which scrapes pitchf/x data from MLB and loads it into the database of your choice using SQLAlchemy. Schedule: 6:00 p.m - 6:30 p.m is time for social. Seminar will start at 6:30 p.m. Please note the office security needs to check your government issued ID (Driver License or Passport, please note Student ID is not valid for entry) . Thanks for your understanding. Our Sponsor: Arity ( https://www.arity.com/ )

  • Tropical Geometry of Deep Neural Networks

    Abstract: Recently Lek-Heng Lim et al., established for the first time, connections between feedforward neural networks with ReLU activation and tropical geometry. In our talk, we aim to explain some of the key results from that paper. We start by introducing tropical algebra concepts such as tropical monomials, tropical polynomials, and tropical newton polygon etc. We then establish the equivalence between neural networks and tropical rational functions. We conclude by , using an example, illustrating how tropical algebra can be used to establish bounds for the complexity of a neural network and to study the decision boundary of a neural network. Bio: Meg Walters is a data scientist and manager at Allstate. She is currently leading a team responsible for researching and prototyping future telematics product offerings for the Allstate brand. Since joining Allstate, Meg has supported telematics initiatives and products including telematics-based pricing, pay per mile insurance, and consumer facing telematics services. She received her PhD in Mathematics from the University of Rochester. Venu Tammali is a data scientist at Arity. Since joining Arity, he has been part of collision detection and distracted driving efforts. Before joining Arity, he worked in a couple of financial firms as a quant. He has a PhD in Mathematics and Bachelors in Electronics Engineering. Schedule: 6:00 p.m - 6:30 p.m is time for social. Seminar will start at 6:30 p.m. Please note the office security needs to check your government issued ID (Driver License or Passport, please note Student ID is not valid for entry) . Thanks for your understanding. Our Sponsor: Arity ( https://www.arity.com/ )

    4
  • Image Caption Generation using Deep Learning in Keras

    Abstract: Image Caption Generation as the name suggests is a method of generating a cogent textual caption for a given image. It is an interesting problem because of its potential applications across domains, but also challenging because of the complexity in terms of the data and models involved. In large part, this talk will focus on a deep learning architecture that I developed from scratch (as a part of a project during my Masters) using Keras comprising an intersection of the fields of Computer Vision and Natural Language Processing. Lately, deep learning methods have achieved state-of-the-art results on examples of this problem, and have become popular for their ability to solve this complex problem without requiring sophisticated data preparation or a pipeline of specifically designed models. In this Python-based talk, we will walk through some of the techniques for simultaneously handling image data with text data as well as touch-base on running your own DL model using Keras via Google Colab (“the free GPU service”). Bio: Swapnil Parkhe is a student at University of Illinois at Chicago where he is pursuing his Master’s degree in Business Analytics (with a focus on machine learning and deep learning). He is also an active Kaggler and AI-meetups attendee, and also works on data-science based projects via various forums. Last summer, Swapnil worked with Akamai as a Business Analyst intern wherein he helped in improving their network deployment processes using data-driven algorithms. Previously, he has worked as a Consultant and Decision Scientist with Equifax and Mu Sigma, respectively across the finance and retail divisions. Schedule: 6:00 p.m - 6:30 p.m is time for social. Seminar will start at 6:30 p.m. Please note the office security needs to check your government issued ID (Driver License or Passport, please note Student ID is not valid for entry) . Thanks for your understanding. Our Sponsor: Arity ( https://www.arity.com/ )

    2
  • SimDivvy - A code first deep dive into simulating Divvy usage across Chicago

    Abstract: Deep dive into an animated and interactive simulation of Divvy bike usage across Chicago developed for a competition. Participants will be exposed to: - Animation using the canvas tag - Concurrent software architecture using message-based multi-process concurrency in Javascript - Developing and using probability distributions to drive the simulation - How interactive “what-if” scenarios were made possible - How this was delivered in a week and how we balanced technical debt vs. value delivered Bio: Justin Bozonier is a Sr. Data Scientist at Arity and has worked in data science for 7 years. He was a lead data scientist at GrubHub and recently published a book titled “Test-Driven Machine Learning”. Justin has over a decade of experience in software development and a BA in Mathematics. In his spare time, he is an Operations Research grad student at GA Tech, a husband/father, and plays the game Skyrim when he finds time (usually in between semesters at school). Twitter: @databozo Blog: http://www.databozo.com Schedule: 6:00 p.m - 6:30 p.m is time for social. Seminar will start at 6:30 p.m. Please note the office security needs to check your government issued ID (Driver License or Passport, please note Student ID is not valid for entry) . Thanks for your understanding. Our Sponsor: Arity ( https://www.arity.com/ )

    1
  • Why Learn PySpark ?

    Arity

    Abstract: Grubhub has chosen to adopt the Spark Big Data computing framework to underpin it’s internal Grubhub Data Platform. Spark was adopted very early by Silicon Valley FANG companies. What features make Spark a great computing platform for both Analytical reporting and Machine Learning? And tips on how to install PySpark on a Mac OSX system so one can play with PySpark without paying for a cloud cluster. Bio: David Liao is currently a data engineer at Grubhub building enterprise digital assets for all. David hold's a Master’s of Data Science from Depaul University. He previously had 12+ years as a Senior Business Intelligence Architect / Developer. Schedule: 6:00 p.m - 6:30 p.m is time for social. Seminar will start at 6:30 p.m. Please note the office security needs to check your government issued ID (Driver License or Passport, please note Student ID is not valid for entry) . Thanks for your understanding. Our Sponsor: Arity ( https://www.arity.com/ )