# MIRI course list study pairs

In­spired by: On learn­ing difficult things

In his re­cent post, user So8res says his num­ber one piece of ad­vice for learn­ing some­thing difficult is to have study part­ner to learn with you.

Since there is a de­cent amount of in­ter­est here in go­ing through the MIRI course list, it might be worth find­ing other peo­ple here to learn and study this with, and to form pairs or groups.

So here is a space for find­ing and or­ga­niz­ing such part­ner­ships!

Of course, part of the rea­son I wrote this is be­cause I am in­ter­ested in learn­ing these books with peo­ple. My back­ground: I’m cur­rently a sec­ond year Ph.D. stu­dent in math­e­mat­ics (num­ber the­ory). I’m still pretty new to the type of math em­pha­sized here. I have Prob­a­bil­is­tic Graph­i­cal Models, Cat­e­gory The­ory for Com­puter Scien­tists and The Logic of Prov­abil­ity (by Ge­orge Boolos—not on the course list, but good to get back­ground for the Ro­bust Co­op­er­a­tion pa­per and for un­der­stand­ing Loeb’s the­o­rem) all ly­ing around. I’m also tak­ing a class on nu­mer­i­cal anal­y­sis. Part of my prob­lem is that I start lots of pro­jects and then end up fiz­zling out on them, and I hope hav­ing a part­ner will help with this.

I’ve already been go­ing through MIRI’s pub­li­ca­tions with a friend from the lo­cal LW com­mu­nity, which has been re­ally nice. I’m still in­ter­ested in find­ing more part­ners <in­sert poly joke here> for go­ing through books on the course list speci­fi­cally. I’m also will­ing to ex­plain things I un­der­stand, or let some­one ex­plain things to me (I’ve found that ex­plain­ing things to some­one else is a very good way of solid­ify­ing your un­der­stand­ing of some­thing) when I have time.

Some things to con­sider:

• What is the best on­line for­mat for do­ing this? I’ve been do­ing this sort of thing with Work­flowy + Math­flowy but there is prob­a­bly some­thing bet­ter.

• Does a pair dy­namic, or a group dy­namic seem more likely to work? I’m hop­ing that there can be a col­lec­tion of pairs all cen­tered in a com­mon com­mu­nity, or some­thing like that.

• If a cen­tral com­mu­nity seems like a good idea, how should it be cen­tral­ized?

• Prob­a­bly some other is­sues/​meta stuff.

• Awe­some.

I am, of course, in the mar­ket for a study part­ner. Please don’t be in­timi­dated by the fact that I’ve been at this for a cou­ple months already: my for­mal ed­u­ca­tion in­cluded bach­e­lors de­grees in com­puter sci­ence and eco­nomics, so I still feel be­hind all you math ma­jors and grad­u­ate stu­dents out there.

I’m not sure how best to make this work, but here are a few of my thoughts:

I recom­mend against limit­ing study­ing to times when you are to­gether. I tend to do a fair bit of study­ing on bus rides and over meals, and I feel that only study­ing to­gether would be some­what limit­ing. It might be use­ful to have an IRC /​ video hang­out go­ing in the back­ground or some­thing, but for me, at least, the ac­tual read­ing of the text is a more pri­vate en­deavor.

I think that most of the gains from pairing up will come from so­cial in­cen­tives (agree­ments to read a cer­tain amount each day/​week, check­ups, etc) and the abil­ity to have con­ver­sa­tions along these lines:

``````So8res > Hey, ping me when you’ve read past the proof at the end of p480
So8res > How is [step] jus­tified? Isn’t that in­valid be­cause [mi­s­un­der­stand­ing]?
Part­ner > Nope. [rea­son]
``````

The abil­ity to have these con­ver­sa­tions would have in­creased my ve­loc­ity pretty sig­nifi­cantly. There doesn’t even need to be real-time com­mu­ni­ca­tion: the abil­ity to email some­one with heav­ily con­tex­tu­al­ized ques­tions, know­ing that they’ve been work­ing on the same con­tent re­cently, would be in­cred­ibly valuable.

I also ex­pect there’s benefit to be had by meet­ing up at the end of each chap­ter and do­ing the fol­low­ing:

1) Mo­ti­vate the ma­jor con­cepts of each chap­ter to each other. There’s a gap be­tween “I un­der­stand what was said” and “I un­der­stand why that was im­por­tant to say”, and ex­plain­ing con­cepts re­ally helps me cross that gap. I’d sug­gest either

• Each mem­ber drafts a brief overview of these con­cepts in, say, an email; you’re not al­lowed to read the other per­son’s email un­til you’ve sent your own.

• You divy up the the­o­rems and each of you ex­plain half to the other.

The first is prob­a­bly more in­ter­est­ing (I’d en­joy con­trast­ing the in­ter­pre­ta­tions), but the lat­ter re­quires half the effort.

2) Do the prob­lems to­gether. As long as both par­ties are par­ti­ci­pat­ing, this may speed things up sig­nifi­cantly. You have to be care­ful to avoid the sce­nario where one party watches while the other solves prob­lems, though. At the very least, there’s benefit to be had by do­ing prob­lems and com­par­ing an­swers.

Th­ese are my best guesses for how to make study­ing eas­ier. Keep in mind that this prob­a­bly varies from per­son to per­son, and that these views are likely to change in ac­tual prac­tice. That said, if this sounds de­sir­able to you, let me know and we can get some­thing started.

• The abil­ity to have these con­ver­sa­tions would have in­creased my ve­loc­ity pretty sig­nifi­cantly. There doesn’t even need to be real-time com­mu­ni­ca­tion: the abil­ity to email some­one with heav­ily con­tex­tu­al­ized ques­tions, know­ing that they’ve been work­ing on the same con­tent re­cently, would be in­cred­ibly valuable.

This is es­pe­cially im­por­tant IMO. Used to be when I stud­ied more ac­tively, when­ever I’d get stuck and some­one I knew hap­pened to be on­line and also study­ing the sub­ject (or had mas­tered it), I’d just ping them up on chat.

• I’m cur­rently a sec­ond year Ph.D. stu­dent in math­e­mat­ics (num­ber the­ory).

My in­ter­pre­ta­tion of your LW user­name just changed abruptly.

• Non-math­e­mat­i­cian here. Just cu­ri­ous… What are you al­lud­ing to? (“Adele_L” sounds like a given name plus an ini­tial to me.)

• It is in fact my name + ini­tial.

But since I am a num­ber the­o­rist with this name, I know that gjm is think­ing adele rings and L-func­tions.

• Yup. (And of course I’d pre­vi­ously as­sumed, cor­rectly as it turns out, that it was given name + ini­tial.) What a nice co­in­ci­dence!

• 12 Nov 2013 10:38 UTC
5 points

I’ve got a copy of Con­cep­tual Math­e­mat­ics: A First In­tro­duc­tion to Cat­e­gories, and I’m more ac­tively work­ing through Ben­jamin Pierce’s Soft­ware Foun­da­tions (which cov­ers a range of ma­te­rial from the Func­tional Pro­gram­ming and Au­to­mated Pro­gram Ver­ifi­ca­tion slots). I’ve got a nice tor­rent full of text­books for var­i­ous branches of math, ma­chine learn­ing, and AI (ex­tend­ing from un­der­grad In­tro to AI up through Hut­ter’s Univer­sal AI).

I can also firmly say that statis­tics class (which I’m tak­ing offi­cially this semester) is a bitch due to the sheer amount of ex­tra sym­bol­ogy to learn and, of course, loads and loads of in­te­grals.

• Name: Eli

• Race: Dwarf

• Class: Spiral War­rior (also sec­ond-year re­search grad-stu­dent)

• Level: two pa­pers in the works, one un­der sub­mis­sion to a conference

• Op­ti­miza­tion Power: I haven’t pre­cisely counted, but my life is very improbable

• The prospect of be­ing for­mally in a study pair/​group makes me anx­ious in case I’m a flake and feel like I’ve be­trayed the other par­ti­ci­pant(s) by be­ing akratic or be­ing un­able to keep up and then I will for­ever af­ter be known as That Flake Who Couldn’t Hack Model The­ory That Every­body Should Laugh At etc. etc. I should prob­a­bly work on that anx­iety, but in the in­terim, as a more pas­sive op­tion, I’ve just cre­ated this Face­book group. Has the benefit that any­body who stum­bles across it or this com­ment can join and dip in at their leisure.

I don’t re­ally know what to ex­pect from the group and I’m fairly con­tent at this point to let its di­rec­tion be driven by who­ever joins, but I would say that if you’re un­sure and hes­i­tat­ing whether to join or post a ques­tion or what­ever, please Just Do It, rather than hov­er­ing, timing out, and giv­ing up. Even if you’re just cu­ri­ous or think you might want to join the group in fu­ture to com­ment but don’t right now, feel free to join now and turn off no­tifi­ca­tions from the group to elimi­nate the Triv­ial In­con­ve­nience for your fu­ture self.

Also, please do feel free to join if you’re not ac­tively study­ing FAI but want to help oth­ers!

• I’m in­ter­ested in study­ing the lat­ter chap­ters of Prob­a­bil­is­tic Graph­i­cal Models, es­pe­cially the parts on struc­ture learn­ing, learn­ing Markov net­works, and de­ci­sion mak­ing.

• I’ve heard of other LessWrongers us­ing Google Han­gouts for this and var­i­ous per­sonal pro­jects, with all kinds of rules or var­i­ants fished from var­i­ous GTD-re­lated strate­gies.

Per­son­ally, I’d be in­ter­ested in set­ting up an IRC cha­t­room, google hang­out or other semi-per­sis­tent thinghy for pair/​group-study­ing of… well, I’m not sure what yet. I’m look­ing at the courses in the MIT Challenge, read­ing through the first book of Feyn­man’s Lec­tures (since it’s fi­nally available for free on­line here, which is the only place I’ve ever found it for less than 200\$ /​w ship­ping), other FAI/​MIRI course, or what­ever else ends up on top of the list af­ter my cur­rent round of pri­or­ity-re­al­ign­ment is over.

• I sec­ond the vote for an IRC room. I spend enough time on IRC already that it would be con­ve­nient.

• Per­son­ally, I’d be in­ter­ested in set­ting up an IRC cha­t­room, google hang­out or other semi-per­sis­tent thinghy for pair/​group

That would be great!

• What’s the cur­rent sta­tus of this? I’m look­ing to get started on the course list and would love a study part­ner.

• From my ex­pe­rience do­ing group study for classes, there don’t seem to be any ma­jor ad­van­tages or dis­ad­van­tages for pairs vs small groups. The most rele­vant fac­tor is how many eye­balls look­ing at some­thing, but even that isn’t a huge effect. Both are more effec­tive than work­ing alone (as the ar­ti­cle con­cludes).

For a lot of things, get­ting to­gether IRL looks like it would work best, but the lo­gis­tics there can be difficult. For peo­ple who have Less­wrong mee­tups nearby, those are an ob­vi­ous way to po­ten­tially co­or­di­nate meatspace study groups.

• I’ve got Cat­e­gories for the Work­ing Math­e­mat­i­cian by Mac Lane; I will be go­ing through this be­cause I will be giv­ing some talks on cat­e­gory the­ory to the math club here at my uni­ver­sity. I pretty much don’t have any logic and I want logic. I have En­der­ton’s A Math­e­mat­i­cal in­tro­duc­tion to logic which is ok, though I think I want to find a new book. I also have Prob­a­bil­ity: The Logic of Science that I want to work through. I also want to go through MIRI pa­pers. I am a math un­der­grad.

I would like to be a part of a study pair or a study group. There seems to be enough peo­ple that we can group to­gether. I would like to learn from peo­ple, and teach peo­ple what I know (mostly pure math: cat­e­gory the­ory/​ab­stract alge­bra/​alge­braic topol­ogy and ba­sic calcu­lus/​real anal­y­sis).

• I am not in­tend­ing on ex­haus­tively go­ing through the MIRI course list, but there is sig­nifi­cant crossover with my study plans. Other com­ments in this post have sug­gested some sort of fo­rum /​ list /​ hang­out around these en­deav­ours. While I can’t com­mit to study­ing alongside a part­ner, I would be en­thu­si­as­tic about par­ti­ci­pat­ing in such a group.

• My copy of Discrete Math­e­mat­ics and Its Ap­pli­ca­tions by Ken­neth Rosen (which is also the text used for the Discrete Maths course re­quired for CS stu­dents at my uni­ver­sity) is sched­uled to ar­rive later to­day. I plan on do­ing an ex­haus­tive read-through (in­clud­ing all or most of the ex­er­cises). I an­ti­ci­pate be­ing able to spend 3 or more hours per week speci­fi­cally on this text. If any­one wants to part­ner up, I’m game.

• I am look­ing for a study part­ner who wants to Ma­chine learn­ing con­cepts with me and per­haps even work on some mini pro­jects to­gether. It should be some­one who is se­ri­ous about de­vel­op­ing Data Science and ML en­g­ineer­ing skills. At the mo­ment, I’m go­ing through Ja­son Brown­lee’s ex­cel­lent (“ad­vanced”) primer on LSTMs as well as read­ing Kag­gle threads on LSTM clas­sifiers that ranked highly in the Kag­gle com­pe­ti­tion. What I’m miss­ing is fel­low learn­ers will­ing to hash through the ma­te­rial to dis­cuss and clar­ify ideas. Want to join me? Feel free to con­tact me at: john.strong@ether­pros.com (john dot strong at ether­pros dot com)(web­site: www.ether­pros.com).

• I’m look­ing for a part­ner to read, study and do the ex­er­cises of the man­ual “Ma­chine Learn­ing in Ac­tion” by Peter Har­ring­ton in ap­prox­i­mately 8 weeks.

• I’m go­ing through the PGM coursera class (It’s one of the classes in the MIRI course-list). I’m definitely go­ing to finish it be­cause I’m do­ing it as an in­de­pen­dent study at my Univer­sity.

Mes­sage me if you’d like to join me. I have a few friends at school who read LW who said they’ll prob­a­bly join me. The more the mer­rier.