EPTThis is a blog about graduate student life before a paper submission.
Last time I wrote about 10 days before a deadline, however, my advisor decided we should submit to a different venue and the deadline is pushed back to 9 / 15. However, I am due to visit China soon (8/13 I fly to San Francisco, stay there for a few days, and go to China for 2 weeks). My adviser demands that I have a rough draft before that day.
Day 0
I have until 8 / 13, so that's just about 8 days starting tomorrow. The algorithm I have seems to work perfectly, except when the search space becomes large. My computer is running out of memory. The root of the problem is I am doing refinement too eagerly, exploring into spaces that really should be ignored (in hind sight). The catch is there is no good way prior to exploring those spaces to know that they are actually wasteful.
So the algorithm needs to be changed in a way that's similar to back to the future. I will need to explore the space with a budget (memory), and when the budget is filled up, I need to go back in time and see what choices I made and if I can discard some of them. By pruning away those unnecessary choices, I am saving spaces on my computer which can then be used to explore spaces that are still of interest (going forward in time).
Now my formerly very simple algorithm needs to be drastically changed. I've been thinking for the past few days on how to properly go back and discard some of the used states, and go forward and I have been unsuccessful at doing so. I'm not overtly worried because whatever the solution may be, it is decidable logic (i.e. not like some really difficult theoretical problem where people are not sure if it is even solvable). Tomorrow I am meeting my adviser (2pm) and I have not yet had a definitive solution. He will also expect some form of rough draft for tomorrow.
Here is an excerpt of my scratch work trying to figure out what's going on. I think my brain is not enough (very often the case in grad school).
![[image loading]](http://i.imgur.com/0WDLZKT.jpg)
The right approach is to rest for a bit, and push yourself over the top. It is like dashing up a hill, you need to do it in 1 shot with momentum. Gradually trying to climb up a slippery hill is not going to cut it. You need to save up and go for a GG push up the ramp.
So I think I'll just take a break, and take it home and continue until I have to sleep. I'll exclusively force myself to write the rough draft tomorrow.
I've been trying this new thing where I use all my computer to play some game and browse internet in the lab before heading home. Frequently I stay up watching stupid youtube videos and/or playing dota. By leaving the computer at the lab I think I'll be able to sleep at a more reasonable time. Hope it works out.
wish me luck
