- AnyLogic: The Platform for Simulation Modeling and State-of-the-Art AI App...
Abstract: In this talk, we will discuss how you can leverage unique features of AnyLogic simulation to solve your business challenges. Aside from demonstrations of the latest development in conventional general-purpose simulation modeling, we will discuss the state-of-the art applications of simulations to train and test AI for business applications. For more information about the topics that we will cover, please refer to: https://www.anylogic.com/features/artificial-intelligence/ Speaker: Arash Mahdavi Simulation/AI Program Lead Dr. Arash Mahdavi is a simulation modeling expert and head of training at The AnyLogic Company in North America. He holds a PhD degree in civil engineering from Purdue University where he applied a system-of-systems approach and agent-based modeling to profitability analysis of construction companies. He has trained hundreds of professionals and faculty members from Fortune 100 companies and elite research universities. He is also currently the AnyLogic-AI integration lead. Tyler Wolfe-Adam Simulation Modeling Specialist Tyler Wolfe-Adam has been a program support specialist at AnyLogic since 2018. His focus has been on providing technical support and helping to advance the AnyLogic AI initiative. He was a co-author of the AnyLogic white paper on AI and simulation and the developer of the Pypeline library. Prior to joining AnyLogic, he received his degree in Computer Science from DePaul University. Changed: Join from a PC, Mac, iPad, iPhone or Android device: Please click this URL to join. https://zoom.us/s/95697416110?pwd=ZVYrdUdBRmFGVnpWM2pZV2hRWEtSQT09 Passcode:[masked] Or join by phone: Dial(for higher quality, dial a number based on your current location): US: [masked] or [masked] or [masked] or [masked] or [masked] or [masked] Webinar ID:[masked] International numbers available: https://zoom.us/u/aRhEyhRs8 Or an H.323/SIP room system: H.323:[masked] (US West)[masked] (US East)[masked] (India Mumbai)[masked] (India Hyderabad)[masked] (Amsterdam Netherlands)[masked] (Germany)[masked] (Australia)[masked] (Singapore)[masked] (Brazil)[masked] (Canada)[masked] (Japan) Webinar ID:[masked] Passcode:[masked] SIP:[masked]@zoomcrc.com Passcode:[masked]
- PyData Global 2020 Online Conference
PyData Global is the very first fully-online PyData conference. Join our global community for five days packed with talks, tutorials, posters, open-source sprints and a digital hallway track.The tickets for the conference are now live and we have exclusively introduced the pay what you can pricing model for everyone attending without the sponsorship of their employer with a minimum ticket price of $10.00 USD.To get your ticket, visit global.pydata.org/pages/tickets
- Introduction to Elyra: Set of AI-centric extensions to JupyterLab
Abstract: Whether you are just getting started in Data Science or are seasoned data scientist, the JupyterLab IDE is likely a tool you are using frequently to get work done. In this session we will introduce Elyra - a set of AI-centric extensions to JupyterLab - that provide support for ML workflow pipelines, Git versioning, code snippets and much more. We'll also demonstrate how to create Machine Learning pipeline from Jupyter notebooks or Python scripts using Elyra's Visual Pipeline Editor, and how to run pipelines locally in JupyterLab or remotely on Kubeflow Pipelines. Short Bio: Upkar Lidder is senior software engineer with 10+ years in IT development including team management, functional and technical leadership roles with a deep experience in full stack technology. Currently focused on Machine Learning and Artificial Intelligence. He can be seen speaking at various conferences and participating in local tech groups and meetups. Upkar went to graduate school in Canada and currently resides in the United States. Patrick Titzler is a developer advocate with the Center for Open Source Data and AI Technologies at IBM. For the past couple of years, he has contributed to several Data and AI open source projects, such as Elyra (a set of AI centric extensions to JupyterLab), the Data Asset Exchange (a curated collection of open data sets for the enterprise), and the Model Asset Exchange (a curated collection of open source deep-learning models). In his previous role as an advocate for IBM's cloud data platform, Patrick focused on open source based data services, such as Apache CouchDB and Apache TinkerPop. Join from a PC, Mac, iPad, iPhone or Android device: Please click this URL to join. https://zoom.us/j/92228985489?pwd=a1N0ZzQ0WWN1dzF6R3F4ZnZtYm9WUT09 Passcode:[masked] Or join by phone: Dial(for higher quality, dial a number based on your current location): US: [masked] or [masked] or [masked] or [masked] or [masked] or [masked] Webinar ID:[masked] International numbers available: https://zoom.us/u/ac9KMltvtQ
- Soft Bayesian Decision Trees
Abstract Neural Networks and Decision Trees: two popular techniques for supervised learning that are seemingly disconnected in their formulation and optimization method, have recently been combined in a single construct. The connection pivots on assembling an artificial Neural Network with nodes that allow for a gate-like function to mimic a tree split, optimized using the standard approach of recursively applying the chain rule to update its parameters. Yet two main challenges have impeded wide use of this hybrid approach: (a) the inability of global gradient ascent techniques to optimize hierarchical parameters (as introduced by the gate function); and (b) the construction of the tree structure, which has relied on standard decision tree algorithms to learn the network topology or incrementally (and heuristically) searching the space at random. Here we propose a probabilistic construct that exploits the idea of a node’s unexplained potential (the total error channeled through the node) in order to decide where to expand further, mimicking the standard tree construction in a Neural Network setting, alongside a modified gradient ascent that first locally optimizes an expanded node before a global optimization. The probabilistic approach allows us to evaluate each new split as a ratio of likelihoods that balances the statistical improvement in explaining the evidence against the additional model complexity — thus providing a natural stopping condition. The result is a novel classification and regression technique that leverages the strength of both: a tree-structure that grows naturally and is simple to interpret with the plasticity of Neural Networks that allow for soft margins and slanted boundaries. Speaker Bio Llu´ıs Antoni Jim´enez Rugama is an algorithmic trading analyst at UBS. Since he joined the company in 2017, he develops algorithms to improve the trading execution in FX markets. Llu´ıs Antoni graduated from Illinois Institute of Technology with a PhD in Applied Mathematics where he focused on error estimation for quasi-Monte Carlo methods under the supervision of Prof. Fred Hickernell. Prior to his PhD, he completed a master’s degree in Financial Engineering in the Ecole des Ponts ParisTech. Join Zoom Meeting https://zoom.us/j/95154074373?pwd=V29zN3FnTVdnZlRiNEkzcHdRZHRpZz09 Meeting ID:[masked] Passcode:[masked]
- QMCPy: A Quasi-Monte Carlo Community Software in Python 3
Abstract Quasi-Monte Carlo (QMC) methods are used to approximate multivariate integrals or expectations of random variables with complex distributions. We have created a Python QMC framework, QMCPy (https://qmcsoftware.github.io/QMCSoftware), that has five main components: a discrete distribution, an integrand, its associated measure, stopping criterion, and summary output data. Information about the integrand is obtained as a sequence of values of the function sampled at the data sites generated by the discrete distribution. The function values are averaged with chosen weights as an estimate of the integral. The stopping criterion computes the error bounds of the QMC estimates and tells the algorithm when a user-specified error tolerance has been satisfied, or to increase the number of sampling points in the next iteration. QMCPy allows researchers and collaborators in the QMC community to develop plug-and-play modules in an effort to produce more efficient and portable QMC software and applications. Each of the aforementioned components is an abstract class, which specifies the common properties and methods of all subclasses. The principal ways in which the five kinds of classes interact with each other are also defined. Subclasses then flesh out different integrands, sampling schemes, and stopping criteria. Besides providing developers a way to link their new ideas with those implemented by the rest of the QMC community, we also aim to provide practitioners with state-of-the-art QMC software for their applications. This is joint work with Fred Hickernell, Sou-Cheng Choi, Michael McCourt, and Jagadeeswaran Rathinavel. Speaker Bio Aleksei is a final-year student at Illinois Institute of Technology working towards a Bachelors in applied mathematics and Masters in data science. He has experience building mathematical software packages and data analysis tools in Python. Join Zoom Meeting https://zoom.us/j/94757811249?pwd=SXJ2QytoYVdkS3NMb0MyaXEzZXZwQT09 Meeting ID:[masked] Passcode:[masked]
- Data Science Generalist
Abstract: There is a lot more to data science than ROC curves. This talk will cover the little things that add up to being a productive data scientist: cleaning data, writing a good PR, building APIs, deployment, communicating with stakeholders, and choosing the right model. Having a variety of basic tools at your disposal like a swiss army knife is key to being an effective data scientist. Speaker Bio: In early 2020 Rory joined the manufacturing startup Fast Radius as a data scientist. He is based in Chicago but spends most days inside except when he's outside running. In 2018 Rory decided to move from Silicon Valley to Linz, Austria to work at Runtastic. At Runtastic (now Adidas Runtastic) he was a data engineer working on data pipelines feeding internal analytics as well as product features. Prior to joining Runtastic, Rory Hartong-Redden was a data scientist at Allstate in Menlo Park, CA. While at Allstate he collaborated with the Stanford Intelligent Systems Lab on autonomous vehicle safety. The project was described in a recent press release and has resulted in several publications. During the winter quarter of 2015 he was a machine learning fellow at Startup.ML where he built a data pipeline to Bitcoin data and set up a platform to backtest trading algorithms. Unfortunately, he did anticipate Bitcoin blowing up the way it did. Rory earned two bachelor’s degrees in Physics and Mechanical Engineering at UC Santa Barbara. His undergraduate research studied the structure of splashed using stereo high-speed cameras. Rory completed a masters in Mechanical Engineering and continued to work in experimental fluid dynamics where he built a new fluid dynamics experiment that measured fluid motion using a novel computer vision system. Zoom Meeting: https://zoom.us/j/94501529558?pwd=a2VBdmxsQzZUNmMycTBORW5EcC81UT09 Meeting ID:[masked] Password:[masked]
- Removing Unfair Bias in Machine Learning
We appreciate IBM's sponsorship to this upcoming meetup. This meetup with be host online. You will receive the event Zoom link after RSVP. Abstract: Extensive evidence has shown that AI can embed human and societal bias and deploy them at scale. And many algorithms are now being reexamined due to illegal bias. So how do you remove bias & discrimination in the machine learning pipeline? In this webinar you'll learn the debiasing techniques that can be implemented by using the open source toolkit AI Fairness 360. AI Fairness 360 (AIF360, https://aif360.mybluemix.net/) is an extensible, open source toolkit for measuring, understanding, and removing AI bias. AIF360 is the first solution that brings together the most widely used bias metrics, bias mitigation algorithms, and metric explainers from the top AI fairness researchers across industry & academia. In this meetup you'll learn: How to measure bias in your data sets & models How to apply the fairness algorithms to reduce bias How to apply a practical use case of bias measurement & mitigation Speaker Bio: Trisha Mahoney Sr. AI Tech Evangelist IBM Trisha Mahoney is an AI Tech Evangelist for IBM with a focus on Fairness & Bias. Trisha has spent the last 10 years working on Artificial Intelligence and Cloud solutions at several Bay Area tech firms including (Salesforce, IBM, Cisco). Prior to that, Trisha spent 8 years working as a data scientist in the chemical detection space. She holds an Electrical Engineering degree and an MBA in Technology Management.
- Machine Learning for Computational Fluid Dynamics
This is an online meetup. Event access: https://zoom.us/j/94294052324 Abstract: Traditionally, scientific computing for analyzing fluid flows has relied on numerically solving nonlinear partial differential equations (PDEs). Unfortunately, these suffer from large computational costs due to the requirement of very fine grids in space and time. Reduced-order models promise the alleviation of these costs by minimizing (or avoiding) the use of PDEs during the forward solve of the fluid flow configuration. This talk will outline our research in the development of such models with highlights such as the use of auto-encoders for nonlinear embedding identification, the use of recurrent neural networks for temporal dynamics evolution and the subsequent construction of inexpensive surrogate models to the PDEs. Moreover, we will present preliminary results from the use of probabilistic modeling methods for quantifying uncertainty in the presence of imperfect data. Our examples will be drawn from many different domains where the understanding and forecasting of fluid flow are critical such as engineering and geoscience. We shall conclude with results from some efforts to bridge the gap between the data science and the high-performance computing software ecosystems here at Argonne National Laboratory. Results shown in this talk were generated from open-source Python packages such as TensorFlow, Numpy, NetCDF4, and Matplotlib in addition to computational physics packages such as OpenFOAM, Drekar, and Paraview. Speaker Bio: Romit Maulik is a postdoctoral researcher at Argonne National Laboratory (ANL). He is the recipient of the Margaret Butler Fellowship in Computational Science at the Argonne Leadership Computing Facility (ALCF). He has strong research interests in artificial intelligence, computational fluid dynamics, and high-performance computing. Associated with ANL’s Mathematics and Computer Science division as well as the ALCF, where he began in 2019, his work revolves around the development of machine learning algorithms for a variety of computational physics applications.
- Cardinality estimation using HyperLogLog... (Online Meetup)
This is an online meetup. The Zoom link is https://zoom.us/j/8057306529 Please register the event so you can receive reminder from Meetup. Topic: Cardinality estimation using HyperLogLog with intersection support and Dask parallel computation Abstract: Cardinality is important to many business applications (e.g., counting the number of unique visitors to a website over a given amount of time). Cardinality estimation methods provided by HyperLogLog are a subclass of probabilistic data structures that approximate cardinality using hashing and other techniques internally to quickly answer an array of cardinality-related questions. I’ll be speaking about a Python implementation of HyperLogLog that I’m modifying to work with Dask, a parallel computation package for Python. So far, the modifications have included serialization and adding the ability to get cardinality for intersections (HyperLogLog-proper calculates cardinality for unions only). I’ll be demonstrating how this could be used to quickly find relationships in large datasets and for data visualization dashboards. Speaker Bio: Scott Little is a data scientist working in digital marketing and has also taught Python and data science classes in Chicago. He has a PhD in Physics from the University of Toledo, where he specialized in thin-film photovoltaic solar cells. For fun he enjoys cycling, electronics, and predicting solar power from satellite imagery and ground photometer sensors.
- High Performance Computing is open for business! (Online Meetup)
We will host this meetup online. After registration, you will receive a Zoom link one day before the event. Topic: High Performance Computing is open for business! Making HPC useful and usable for commercial applications in simulation, analytics and machine learning. Abstract: High Performance Computing - HPC - has long been a vital tool for academic and government research. It brings to mind room-sized supercomputers that focus on modeling and analyzing important phenomena ranging in scale from molecules, genes and materials, to vehicles, aircraft, weather and the cosmos. It may surprise you to learn that the Chicago area and the greater Midwest is a mecca of HPC-powered research at national laboratories and universities. Parallel Works is a Chicago-based company whose mission is to make HPC usable by business and industry, enabling companies large and small to easily leverage the benefits of HPC on Cloud and on-premise resources. In this talk, we’ll explain the basics of high performance computing and show many examples of its benefits in fields as diverse as manufacturing, auto racing, transportation, and healthcare. And we’ll highlight how one Chicago company is taking down the barriers to true democratization of this important tool for global progress in fields that matter to our everyday lives. Speaker Bio: Michael Wilde is Co-Founder and CEO of Parallel Works Inc. Since 2001 he served as Software Architect at Argonne National Laboratory (where he is on entrepreneurial leave), and is a research computer scientist at the University of Chicago. Before joining Argonne in 2001, Wilde served as CTO of Chicago-based startup CoolSavings.com, and consulted at AT&T Bell Laboratories in the areas of mainframe UNIX and programming models for large-scale software development.