Commit c7d58530 authored by Mario Surlemont's avatar Mario Surlemont
Browse files

Update todo.md

parent b719a2af
......@@ -89,16 +89,20 @@
- From now on only separators between the root node and other components are added to the model. This improves the runtime. ✔
- The ring inequalites haven't been implemented yet. ❌
- I would prefer to focus on the writing and further testing prior to improve the formulation. If there is enough time left I will try to further improve the runtime.
* May implement that cuts are added even for fractional solutions to strengthen the LP bound. (thursday) ❌ (moved to baacklog=)
*
### runtime
* Test the ASP version on the graphs from week 5 to compare them. (saturday)
* Test the (new) implementation on the usual graphs and them from week 5. (saturday)
* Create clean tables and CSV files. (sunday)
* May implement that cuts are added even for fractional solutions to strengthen the LP bound. (thursday) ❌ (moved to bbacklog)
### runtime
* Test the ASP version on the graphs from week 5 to compare them. (saturday) ✔
- The ASP version performs quite bad on many random graphs where the ILP version performs quite good. One key factor that I could determine is that on those random graphs were the ILP version was good the difference in the size of an unconnected solution and a connected is rather low. So the ILP version seems to handle those instances were there is no big difference good.
- The ASP version performed in general better on grid graphs. This is what I expected as the ASP version was better on the leaf graphs as well. My assumption still is that there are too many "alternative" unconnected solutions where it is much cheaper to replace nodes from an previous unconnected solution and to generate a new unconnected solution than to add those vertices which separate them. As for those instances where the ILP needs a lot of time to solve the number of lazily added constraints is very high and the gap decreases(my test results) very slow both of these facts seem to underline my assumption.
* Test the (new) implementation on the usual graphs and them from week 5. (saturday) (❌ )
- Only tested on a few graphs as the testing costs much time. But the decrease in runtime was significant.
* Create clean tables and CSV files. (sunday) (❌ )
- As I tested many graphs I have to further select which graphs I really want to include and I might have to do new tests if I change the implementation.
## thesis
* Begin the paragraph of results. (wednesday)
* Refactor the paragraph for Implementation and methods. (wednesday)
* Begin the paragraph of results. (wednesday) (✔)
- I wrote those bullet points that I want to include in a text file. But I haven't created an tex file yet.
* Refactor the paragraph for Implementation and methods. (wednesday) ✔
# Backlog
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment