@mistersql The GenAI version involves the same people but instead prompts them with randomly generated assembly and asks them to score it.
Optimization AI — Integer and Constraint Programming. Decision Diagrams. Cat and early music enthusiast. Founding nextmv.io. Prefers Go, Python, and Julia.
@mistersql The GenAI version involves the same people but instead prompts them with randomly generated assembly and asks them to score it.
I've finally had the chance to play with Gurobi's hierarchical optimization features, and they're pretty nice!
https://ryanjoneil.github.io/posts/2024-11-08-hierarchical-optimization-with-gurobi/
As I've been renovating old blog posts, some are more fun that others. This one was kind of fun to update with the latest #SCIP libraries.
https://ryanjoneil.github.io/posts/2012-01-12-magic-squares-and-big-ms/
This looks like a fun new book!
Problems and Solutions for Integer and Combinatorial Optimization: Building Skills in Discrete Optimization
Topics: modeling capabilities of integer variables, the Branch-and-Bound method, cutting planes, network optimization models, shortest path problems, optimum tree problems, maximal cardinality matching problems, matching-covering duality, symmetric and asymmetric TSP, 2-matching and 1-tree relaxations, VRP formulations, and DP.
https://my.siam.org/Store/Product/viewproduct/?ProductId=49696448
It's really neat seeing how active a project #Pyomo is.
This post riffs a bit on a recent post by Erwin Kalvelagen about using #optimization to schedule team meetings, and attempt to reduce double- and triple-booking using #julia, #jump, and #scip.
https://ryanjoneil.github.io/posts/2023-11-26-reducing-overscheduling/
“The application of OR-Tools and LAD regression for forecasting feels new, but the use of LP for shift scheduling is timeless.”
@ryanjoneil demos OR-Tools, HiGHS, Nextroute: https://nextmv.io/videos/forecast-schedule-route-3-starter-models-for-on-demand-logistics
#datascience #Python #logistics #lastmile #linearprogramming #orms #informs #ORTools
At @nextmv we like to do #optimization speed runs. In a couple weeks, we'll be upping the ante and do 3 in 30 minutes.
Come see the 3 main models you need to operate an #ondemand #lastmile service using 3 different solvers: #ORTools, #HiGHS and our own #nextroute.
https://www.nextmv.io/videos/forecast-schedule-route-3-starter-models-for-on-demand-logistics
Guilhem Molines of IBM discusses the value of decision rules over code at #DecisionCAMP 2023:
- transparency
- traceability
- explainability
All important parts of decision #models.
Denis Gagne at DecisionCAMP 2023:
"Symbolic #AI has proven better for decision automation" partly because it is "better at abstract problems."
Sub-Symbolic AI is "better at sensing and perceptual problems."
Gary Hallmark showing decision modeling using #chatgpt at DecisionCAMP 2023 (paraphrased):
"It feels like I'm spending more time on prompt engineering than I spent on software engineering."
Interesting observation: GPT seems to be better at generating useful Python code than JSON data.
Visualizing decision diagrams #izhard.
https://ryanjoneil.github.io/posts/2023-09-13-visualizing-decision-diagrams/
Granted, her grandmother defended in her 60s, so 📉?
Hope she likes her thesis advisor!
This morning I wished my toddler well as she went off to school and realized that, if she is like me, this will be her life for the next 37 years.
Big idea: Name research projects after other researchers’ cats
#Python language gotchas can be pretty crazy. I remember them making it hard to build clean APIs back when I did a lot of Python dev. Here's an example.
Likely our last good beach day of the year in #Maine.