Join the conference exhibitors as they discuss innovations and best practices in the field. Professional Development Units (PDUs) are available to those who attend these sessions.
Descriptions and times below:
Monday, April 17
Time:
9:10-10am
Location:
Cottonwood 11
Decision Model, Meet Production: A Collaborative Workflow for Optimizing More Operations
Time:
9:10-10am
Location:
Cottonwood 10
Machine Learning in Text Analytics: Do We Really Need Deep Learning?
Presented by: Normand Peladeau
The renewed enthusiasm for artificial intelligence (A.I.) and, more particularly, for techniques based on deep learning and other forms of neural networks, means that we are trying to apply these latest techniques to all problems requiring a supervised or unsupervised form of learning. But this unprecedented wave of interest often makes us forget there are other forms of machine learning that have proven themselves over time. During this presentation we will compare certain forms of machine learning with and without the contribution of neural network techniques in order to assess the importance and the nature of a possible contribution (if any). To do this, we will examine different tasks in the field of automatic language processing, namely topic modeling, automatic word disambiguation, and the development of semantic lexicons. We will also try to identify in which context an approach based on neural networks or deep learning deserves consideration.
Time:
10:30-11:20am
Location:
Cottonwood 10
Gurobi’s Newest Educational Resources: Where Data Meets Decisions – An Overview of Our Free Jupyter Notebook Data Science Example Library
Presented by: Rahul Swamy and Jerry Yurchisin
How can you use different prediction models for avocado price optimization? How can you identify plagiarism with text similarity? How can you effectively plan for airline disruption in time of continual flight delays and cancelations? How can you discover lesser-known artists in your daily music playlists? How can you build the perfect fantasy basketball team?
By combining data science tools and mathematical optimization.
In this session, Gurobi will introduce several of our newest (and free) educational examples that students and instructors can use to learn and teach real-world applications of combined data science and optimization problem solving. We will review our new data science library of Python Notebook Examples that combine predictive and prescriptive analytics and offer new data science learners an entry point into problem-solving with optimization.
Time:
11:30am-12:20pm
Location:
Cottonwood 10
Quickly Deploy your Optimization Models to the Cloud with DBOS!
Presented by: Giulia Burchi and Filippo Focacci
DecisionBrain Optimization Server (DBOS) is designed to help build and deploy fully scalable optimization-based applications. It enables optimization developers to focus on their models, benchmark them and allows them to effortlessly deploy those models in production in a context that will support multiple parallel runs on dedicated resources. To achieve this, DBOS lets you encapsulate any computational module (optimization solvers, analytics modules, etc.) into so-called “Workers.” Workers can be deployed on dedicated resources (local, private, or public cloud) to ensure the best execution time. When deployed on Kubernetes, Workers may be activated on-demand to reduce cloud costs. DBOS can be used in a stand-alone mode to run computations, or it can also be integrated with existing applications to let them provide scalable and on-demand optimization capabilities and powerful monitoring capabilities. DBOS also has a benchmarking functionality that allows you to benchmark your optimization engine across versions, different datasets, or models. In this presentation, we will demonstrate how this technology can be used to:
- Encapsulate an optimization model in a Worker
- Deploy this Worker on a Kubernetes cluster using resources only on-demand
- Monitor Real-time Executions
- Benchmark models and datasets
Time:
11:30am-12:20pm
Location:
Cottonwood 11
Quantum Computing for Optimization
Presented by: Alex Koszegi
Quantum computing has gone from the lab to the enterprise, and a recent Hyperion Research study found that that there already are a wide range of commercial organizations engaged in some form of quantum computing efforts. While you may think production use of quantum computers are years away, the first commercial quantum applications are in production are using D-Wave’s quantum technology. During this talk our speaker will discuss how quantum computers can be used to solve complex optimization problems, give examples of relevant use cases, and explain how enterprises can get started on their quantum journey.
Time:
1:50-2:40pm
Location:
Cottonwood 10
ODH Python Primer
Presented by: Robert Ashford
This short tutorial shows participants how to build a basic model using the ODH in Python. This session includes setting the Python environment, reading data from a CSV or spreadsheet, creating variables, objective functions, and constraints, solving the model, and returning the results. Additionally, this session points the participants to further reading so that they may expand their capabilities. Furthermore, we will present the brand-new ODH generic API and demonstrate it in Python (with Links to CPLEX, Gurobi, and FICO XPRESS.
Time:
1:50-2:40pm
Location:
Cottonwood 11
Arena and Emulate3d
Presented by: Nancy Zupick
In this session, we will introduce the Arena and Emulate3d software packages, examine what types of systems they can model and what problems they can help you solve, and discuss training options for both.
Time:
3:40-4:30pm
Location:
Cottonwood 11
Python and AMPL: Build Prescriptive Analytics applications quickly with Pandas, Colab, Streamlit, and amplpy
Presented by: Filipe Brandão and Robert Fourer
Python and its vast ecosystem are great for data pre-processing, solution analysis, and visualization, but Python’s design as a general-purpose programming language makes it less than ideal for expressing the complex optimization problems typical of prescriptive analytics. AMPL is a declarative language that is designed for describing optimization problems and that integrates naturally with Python. In this presentation, you’ll learn how the combination of AMPL modeling with Python environments and tools have made optimization software more natural to use, faster to run, and easier to integrate with enterprise systems. Following a quick introduction to model-based optimization, we will show how AMPL and Python work together in a range of contexts:
- Installing AMPL and solvers as Python packages
- Importing and exporting data naturally from/to Python data structures such as Pandas dataframes
- Developing AMPL model formulations directly in Jupyter notebooks
- Using AMPL and open-source solvers for free on Google Colab, with no arbitrary problem size limits
- Turning Python scripts into prescriptive analytics applications in minutes with Pandas, Streamlit, and amplpy
Time:
3:40-4:30pm
Location:
Cottonwood 1
End-to-End FICO® Xpress Insight Tutorial: From Data to Decisions for Non-Technical Business Users
Presented by: Majid Bazrafshan
Time:
3:40-4:30pm
Location:
Cottonwood 10
DECIDE BETTER: the Decision Science Lifecycle
Presented by: Matt Brady
What is the current state of the art in Decision Science? Attend this interactive workshop to understand the full lifecycle, from pre-mortem to robust decision to post-mortem. Engage as we work through actual audience scenarios via a decision architecture process, and experience the collaborative decision optimization (TM) that the Volley platform enables.
Be sure to attend our immersive Technology Workshop (Sun Apr 16, 3:00 – 4:45 pm) to understand the strategies, motivations, and techniques of Decision Science. Then stop by our Booth (#300) to see the innovative Geography Explorer and API Integration in action, and follow all the progress on LinkedIn. Decide better with Volley.
Tuesday, April 18
Time:
9:10-10am
Location:
Cottonwood 11
Gurobi Machine Learning: Incorporate your Machine Learning Models into Optimization
Presented by: Alison Cozad and Zed Dean
Gurobi is making it easier to plug your predictive models directly into your optimization model. The Gurobi Machine Learning is an experimental, open-source Python package that allows users to add trained machine learning regressors as a constraint to a Gurobi model (e.g., from scikit-learn, TensorFlow/Keras, or PyTorch). Thus, allowing for tighter integration between trained predictions and optimal decision-making.
This tutorial will introduce the Gurobi Machine Learning package and how it fits into an optimization application. Then we will explore how these machine-learning models are incorporated into a Gurobi model through a couple of examples.
Time:
9:10-10am
Location:
Cottonwood 10
End User Responsive Analytics: A Python Lightweight Server Framework
Presented by Irv Lustig, PhD
The end users of many analytics applications want to press the “Solve” button in a browser-based application and get a quick response to their business challenge. For example, an end user may want to spend at most a few seconds to create a production schedule for a business operation, or quickly assign people to jobs. Princeton Consultants has built a lightweight Python framework that simplifies the delivery of the back-end server for such applications. This avoids the complexity of other frameworks that are more suited for applications where the analytics process is computationally expensive. In this tutorial, we will demonstrate our best practices for developing analytics applications in terms of processes, Python libraries, and development tools, using optimization as a motivating example.
Time:
11:30am-12:20pm
Location:
Cottonwood 11
Maintenance Optimization: Presentation of a Data-Driven Predictive Maintenance Planning Framework Project
Presented by: Renaud Saltet
Maintenance strategy is crucial to minimize downtimes and costs and maximize production. Artelys develops optimization solutions for resource scheduling in logistics and transportation. Artelys Crystal Resource Optimizer is a software specialized in resources planning under constraints that supports your company through all the steps of its planning process.
As the amount of available data grows, predictive maintenance has become increasingly effective to detect anomalies and defects in equipment. Artelys is conducting a project on predictive maintenance in which a discrete-event simulator replicates the system at hand and produces scenarios based on the components interdependencies, aging, maintenance operations, and sensitivity to external factors such as weather. Scenarios are used to assess and optimize a maintenance strategy through visualization and KPIs. The goal is to design a robust planning that minimizes the need for curative maintenance, that is, repairing unexpected failures at a high cost.
This tutorial will introduce Artelys resource optimization solutions and dive into concepts and tools from survival analysis to develop a module for maintenance planning optimization that incorporates predictions on the system’s condition.
Time:
11:30am-12:20pm
Location:
Cottonwood 10
Text Mining and Sentiment Analysis to the Curriculum of Introductory Analytics Courses
Presented by: Kevin Potcner
JMP Pro Statistical software has made analyzing unstructured text data simple and engaging. Requiring no prior experience in the concepts of formal statistical analyses (confidence intervals, p-values, models, etc.), extracting meaning from a large collection of text can now be done by even students brand new to the world of analytics.
And with today’s students being intimately familiar with these type of data, the value of such analyses is easily appreciated by any student. Using JMP Pro statistical software, the presenter will illustrate the process of analyzing text data to uncover key themes and quantify responders’ sentiment.
Time:
1:50-2:40pm
Location:
Cottonwood 11
Nonlinear Optimization Using Artelys Knitro
Presented by: Richard Waltz
Nonlinear optimization is used in many applications in areas such as finance, energy, health, 3D modeling, and marketing. With four algorithms and great configuration capabilities, Artelys Knitro is the leading solver for nonlinear optimization and demonstrates high performance for large-scale problems. This session will introduce you to Artelys Knitro, its key features and modeling capabilities, with a particular emphasis on the latest major improvements including recent advances in solving mixed-integer nonlinear optimization problems. We will also provide benchmarks highlighting the power of Knitro to efficiently solve large-scale, nonlinear models with hundreds of thousands of variables and constraints.
Time:
1:50-2:40pm
Location:
Cottonwood 10
Model Deployment and Data Wrangling with GAMS Engine and GAMS Transfer
Presented by: Adam Christensen
The right tools help you deploy your GAMS model and maximize the impact of your decision support application.
GAMS Engine is a powerful tool for solving GAMS models, either on-prem or in the cloud. Engine acts as a broker between applications or users with GAMS models to solve and the computational resources used for this task. Central to Engine is a modern REST API that provides an interface to a scalable, containerized system of services, providing API, database, queue, and a configurable number of GAMS workers. GAMS Engine is available as a standalone application, or as a Software-As-A-Service solution running on AWS.
GAMS Transfer is an API (available in Python, Matlab, and soon R) that makes moving data between GAMS and your computational environment fast and easy. By leveraging open source data science tools such as Pandas/Numpy, GAMS Transfer is able to take advantage of a suite of useful (and platform independent) I/O tools to deposit data into GDX or withdraw GDX results to a number of data endpoints (i.e., visualizations, databases, etc.).
In this session we will go through the necessary steps to get started with GAMS Engine and GAMS Transfer.
