Intro to Optimization Through the Lens of Data Science Pt. 2

Name: Intro to Optimization Through the Lens of Data Science Pt. 2
Rating: 4.8 (171 reviews)

A comprehensive introduction to mathematical optimization and gurobipy tailored to data scientists and problem solvers

Created byDr. Joel Sokol

Last updated 4/2024

English

English [Auto],

What you'll learn

What is optimization and how can it be applied to complex problems?
How to identify an optimization problem and translate real life into optimization models
Learn about solvers and algorithms
Introduction to Gurobi/gurobipy and using it in exercises and real-world problem solving

Course content

4 sections • 13 lectures • 1h 40m total length

Optimization in Data Science8:15
Discover how optimization underpins data science by finding best parameter values for regression, classification, clustering, and time series models.
Customizing Data Science Models Using Optimization8:26
Explore how optimization customizes data science models, adjusting regression loss, adding constraints like lasso, ridge, and elasticnet, and directly selecting a best subset of features with binary variables.

Modeling Example Introduction10:00
Develop an optimization model to minimize jet fuel cost across four cities and learn balance constraints that link purchase, takeoff, and landing fuel for a coherent ground and air segment.
Recursive Looking Constraints8:02
Learn to model airline fuel optimization with recursive looking constraints in data science, linking takeoff, landing, and purchases while respecting tank capacity and safety margins.
Nonnegativity6:19
Explore how non-negativity constraints shape optimization in a jet fuel purchasing model, showing why some variables can be negative and why bounds must be specified.
Variable Substitution and Extra Variables7:07
Learn how variable substitution with extra variables simplifies a jet-fuel optimization model, replacing y with z plus x. Preserve non-negativity constraints and let solver substitution handle size.
How Optimization Questions are Presented7:32
Explore how to turn vague English descriptions into mathematical optimization models for data science, including objectives, constraints, and data-driven iterative discovery with predictive and probabilistic models.
Modeling Note: Quadratic Objectives5:14
Explore efficient quadratic modeling by introducing an auxiliary variable z to represent the sum, then minimize z^2; compare two formulations and extend to quadratic constraints to speed up Gurobi optimization.

Reusability of Models2:50
Read data from files to make optimization models reusable across different days and scenarios. Use the same Gurobi model for many cyclic routes by swapping data files.
Infeasibility and Debugging9:26
Learn to model optimization for diamond procurement, detect infeasibility from strict equalities, and debug by adjusting data and constraints to reveal cheaper supplier mixes.
Using Models to Analyze Sensitivity11:16
Learn how to use optimization models to analyze sensitivity and answer what-if questions, minimizing costs while meeting demand and evaluating capacity upgrades in chemical production.

Classification of Optimization Models10:20
Classify optimization models by their math, from linear programs to mixed integer and non-convex quadratically constrained programs, and learn how solvers like Gurobi handle large and complex instances.
Solution Algorithms at a High Level5:30
Understand how optimization models start from an initial solution and iteratively improve via an improving direction and step size, like gradient boosting and Newton's method, with branching and multiple starts.

Requirements

Basic Python, college-level mathematics, experience with Jupyter Notebooks, and the ability to open python (.py) and Jupyter Notebooks (.ipynb) files on your machine

Description

Welcome to Introduction to Optimization Through the Lens of Data Science!

This free 4-part course was developed to help teach data scientists how to add optimization to their toolbox and when to use it in their advanced problem-solving. We will cover a comprehensive introduction to optimization, when optimization is the best tool to solve a problem, and how to translate real-life problems into optimization.

We will introduce you to world-class tools to help you problem solve, and provide everything from basic hands-on exercises to more advanced full real-world use cases to reinforce all new concepts of prescriptive analytics as you learn them. We look forward to having you learn optimization (and gurobipy) with expertise from Dr. Joel Sokol and the team of Ph.D. experts from Gurobi Optimization, who helped develop this comprehensive introduction to mathematical optimization.

In part 2, you will dive deeper into the relationship between optimization and data science. Work with more complex constraints, understand model reusability, analyze sensitivity, and understand infeasibility. Classify types of optimization problems and see how they are solved at a high level.

Hands-on Exercises:

Please check the resource section of many of the lectures to find self-assessments in the form of exercise files and solution files. You will also notice we have data and code files available to help you work your way through these practice exercises.

Who this course is for:

Data scientists and problem solvers curious about mathematical optimization/prescriptive analytics.

Intro to Optimization Through the Lens of Data Science Pt. 2

What you'll learn

Explore related topics

Course content

Optimization & Data Science2 lectures • 17min

More In-depth Modeling6 lectures • 44min

Reusability, Infeasibility, and Analyzing Sensitivity3 lectures • 24min

Classification of Models and Algorithms2 lectures • 16min

Requirements

Description

Who this course is for: