goddamn advisor

[Gary_Yngve]

Unless I start feeling passionately in love with my project, he's not going to fund me. This, after I've worked nights, weekends, no mountains, etc. for the past three years, and provided my own funding (that just ran out).

 

He insists that I do not have to work more hours, just every hour has to be so productive... and when I'm in the mountains, all I'm thinking about is what I'm going to do next with my project.

 

So for the next few months, I'm supposed to pretend to be passionate about the project, completely ignoring the axe over my head.

 

Oh yeah, and he gave me a new project to work on for which I should have results and a short paper done by Tuesday, after he had earlier denied that he had ever forced me to work weekends.

[skykilo]

That sucks, man. Do you already have a project of your own? If not, then why don't you look for a new advisor? I guess I should ask about your field of research...

[Greg_W]

Didn't you have some problem with some bitch in a position of authority over you at the University last month? Maybe the problem isn't your advisor, maybe it's you. Have you ever thought that maybe you have problems with most authority figures? And if you don't now, you make one up?

[Gary_Yngve]

Greg_W said:

Didn't you have some problem with some bitch in a position of authority over you at the University last month? Maybe the problem isn't your advisor, maybe it's you. Have you ever thought that maybe you have problems with most authority figures? And if you don't now, you make one up?

 

If I had problems with authority figures in general, I wouldn't be taking Mountaineers courses.

 

The bitch at the IMA is a well known troublemaker who has given anyone and everyone a hard time.

[Gary_Yngve]

skykilo said:

That sucks, man. Do you already have a project of your own? If not, then why don't you look for a new advisor? I guess I should ask about your field of research...

 

I haven't had a chance yet to ask around to obtain funding from other profs. Something I probably should do...

 

I told him that I expect to go climbing on the weekends. He told me that if I expect that, I shouldn't be in grad school.

 

[skykilo]

Don't ask, don't tell, man!

Find a cool prof., they're definitely out there...

[klenke]

A good friend of mine who was in the PhD program in the Mechanical Engineering Department at UW was always complaining to me about how her advisor (who, by my own experiences, is/was a complete prick) would keep heaping on her his own little projects thereby not allowing her to get on with her own.

 

I guess this is standard operating procedure for profs: to dump the lame work they don't want to do or can't do themselves on their minions.

 

Sorry to hear about your problem(s). Literally I am sorry, having considered at length going back to school to pursue my own PhD (just not in Computer Science).

[slothrop]

Yeah, man, tell him to fuck off and find someone else. Maybe the "problem" Greg alludes to is the ease with which you give in to your advisor's bullying.

 

If he wants you to be more productive, point out to him that extra projects assigned over a weekend will not help that cause.

 

There's got to be a better professor out there.

Edited August 29, 2003 by slothrop

[Gary_Yngve]

skykilo said:

That sucks, man. Do you already have a project of your own? If not, then why don't you look for a new advisor? I guess I should ask about your field of research...

 

My current project is aimed at the mostly unsolved problem of controller reduction. I started the project six months ago. At least three of those months were spent doing other crap (taking a class, TAing, peer reviewing a few papers, revising a paper accepted to a conference, preparing said talk...). During the three months I had to work, I implemented balanced truncation for model reduction on linear systems, adapted spacetime optimization to work with Galerkin projections, ported a fast forward dynamics engine over to our system, read a bunch of papers, and spent countless hours debugging and tuning nonlinear constrained optimization solvers. But I haven't gotten any compelling results yet, and my advisor thinks it's because I'm not passionate about the problem. My counter is that I cannot be passionate about the problem if I'm feeling threatened to get results.

 

The new project involves using adjoint methods to solve spacetime problems faster, and it may be useful in the project that I have been working on.

[fleblebleb]

You so just killed your thread.

 

Garblegarbleblahblahgarblebling...

[klenke]

No you didn't Gary. I'm bored.

 

So, elaborate on this sentence:

"My current project is aimed at the mostly unsolved problem of controller reduction."

 

What is controller reduction?

And why is it mostly unsolved?

[j_b]

whether or not you should work on weekends depends on the culture of your department. what do other students do? what about your advisor's other students? you know the answer to these questions and if you can't fit in (which is completely allright) perhaps you should look somewhere else for something more adapted to your expectations.

[Attitude]

Gary_Yngve said:

Unless I start feeling passionately in love with my project, he's not going to fund me.

Explain to him that you were brought up in a repressive environment and don't openly express love. You are terribly passionate about your project even though you don't appear to be. Then show him your prescription for anti-depressants.

[lummox]

grow some balls. tell the fucker 'no' and if he starts giving you shit get all aggro. tell him youll get medival on his ass. if you want respect you gotta demand it. know what im saying. otherwise throw your fucking life out the window along with your pride.

[Sphinx]

You should get together with Dryad.

[TIGHE]

That's exactly why I couldn't do the "PHD" thing. I couldn't sell my soul for the fuckin' piece of paper.

 

I stopped at a Master's because it was all the bullshit manipulation I could stomach.

Good luck with finding some funding.

[skyclimb]

I am currently doing some biotech research for a prof. At first I was a little turned off by the amount of hard work it takes, but then I cam to the realization that this work is like being a contractor. You have to get into that mind frame, and not one of a student. That is when I started having a life again. Hope this helps.

[Gary_Yngve]

j_b said:

whether or not you should work on weekends depends on the culture of your department. what do other students do? what about your advisor's other students? you know the answer to these questions and if you can't fit in (which is completely allright) perhaps you should look somewhere else for something more adapted to your expectations.

 

The funny thing is that I am one of his few remaining students. Two students quit on him. Three are co-advised, so they only get half of his crap. One is the stereotypical work-in-the-corner-all-day-and-talk-to-no-one Chinese guy.

[Gary_Yngve]

klenke said:

No you didn't Gary. I'm bored.

 

So, elaborate on this sentence:

"My current project is aimed at the mostly unsolved problem of controller reduction."

 

What is controller reduction?

And why is it mostly unsolved?

 

Suppose you have a humanoid robot. It has many joints, and it has motors on each of the joints. The robot has sensors to detect all sorts of things. How do you make the robot walk?

 

You have to figure out how to control each motor for each joint, based on the current joint positions and all the sensory input that you have.

 

Model reduction focuses on simplifying the state information so that the inputs get mapped as best as possible to the outputs (controls to observations).

 

Control reduction focuses on simplifying the control parameters (knobs/dials for the motors) so that control problems (which tend to be nonlinear and nasty) can be solved in a lower-dimensional (easier) space. Ideally a reduced controller should retain as much flexibility from the full controller and behave robustly.

 

The tricky part about control reduction is that the reduced controllers have to operate on a reduced model. One option is to do model reduction and then design a controller on the simplified model. The problem with this method is that the approximation happens at the beginning, and errors tend to compound. Ideally one would want to simultaneously simplify the model and the controller at the same time (i.e. with a closed loop between model and controller). There are esoteric mathematical methods to do this for linear systems, but extending to nonlinear systems and performing robustly is still largely open.

 

The other way to simplify control is to make control not necessary, that is, make the system inherently robust. A simple example would be a slinky. It tends to walk down stairs no matter what width/height the stairs are. More complicated examples have been studied in biomechanical literature.

 

We've explored both directions, and we've tried working with toy problems (multi-link pendulums) and with real problems (humans walking).

 

No killer results yet.

[Gary_Yngve]

lummox said:

grow some balls. tell the fucker 'no' and if he starts giving you shit get all aggro. tell him youll get medival on his ass. if you want respect you gotta demand it. know what im saying. otherwise throw your fucking life out the window along with your pride.

 

I sent this email to him prior to our fight today...

I either had balls or was damn stupid:

 

I have no retreat paper, unless I can pull results out of my ass.

 

So should I actually enjoy my weekend, or should I work away yet another weekend in hopes that results are going to fall out of the sky and hit me on the head, because not submitting a retreat paper is disappointing and unacceptable?

 

[Gary_Yngve]

Greg_W said:

Didn't you have some problem with some bitch in a position of authority over you at the University last month? Maybe the problem isn't your advisor, maybe it's you. Have you ever thought that maybe you have problems with most authority figures? And if you don't now, you make one up?

 

I had three great relationships with advisors in undergrad.

 

I do think I can be a little passive-aggressive though... not sticking up for myself when I should, and then being angry later...

[Gary_Yngve]

Grrrr, and I had to cancel my plans for this weekend.

 

At least half a dozen times this year I've had to cancel or bail from climbing plans because either my advisor dumped more work on me (or pushed up a deadline) or I've been too exhausted to go.

 

Had dinner tonight with a friend of mine who quit grad school a year ago and is now raking in the big bucks...

decisions, decisions...

[Fence_Sitter]

Gary_Yngve said:

klenke said:

No you didn't Gary. I'm bored.

 

So, elaborate on this sentence:

"My current project is aimed at the mostly unsolved problem of controller reduction."

 

What is controller reduction?

And why is it mostly unsolved?

 

Suppose you have a humanoid robot. It has many joints, and it has motors on each of the joints. The robot has sensors to detect all sorts of things. How do you make the robot walk?

 

You have to figure out how to control each motor for each joint, based on the current joint positions and all the sensory input that you have.

 

Model reduction focuses on simplifying the state information so that the inputs get mapped as best as possible to the outputs (controls to observations).

 

Control reduction focuses on simplifying the control parameters (knobs/dials for the motors) so that control problems (which tend to be nonlinear and nasty) can be solved in a lower-dimensional (easier) space. Ideally a reduced controller should retain as much flexibility from the full controller and behave robustly.

 

The tricky part about control reduction is that the reduced controllers have to operate on a reduced model. One option is to do model reduction and then design a controller on the simplified model. The problem with this method is that the approximation happens at the beginning, and errors tend to compound. Ideally one would want to simultaneously simplify the model and the controller at the same time (i.e. with a closed loop between model and controller). There are esoteric mathematical methods to do this for linear systems, but extending to nonlinear systems and performing robustly is still largely open.

 

The other way to simplify control is to make control not necessary, that is, make the system inherently robust. A simple example would be a slinky. It tends to walk down stairs no matter what width/height the stairs are. More complicated examples have been studied in biomechanical literature.

 

We've explored both directions, and we've tried working with toy problems (multi-link pendulums) and with real problems (humans walking).

 

No killer results yet.

 

damn

[Gary_Yngve]

Attitude said:

Gary_Yngve said:

Unless I start feeling passionately in love with my project, he's not going to fund me.

Explain to him that you were brought up in a repressive environment and don't openly express love. You are terribly passionate about your project even though you don't appear to be. Then show him your prescription for anti-depressants.

 

That's actually half-true... I'm not very open with my emotions.

[lummox]

so hows it feel to be punked out in the prison of 'higher learning'?