I'm currently doing research in pseudorandomness, which involves a zoo of wonderful objects such as pseudorandom generators, randomness extractors, expander graphs, etc. I find it a fascinating topic, but one thing that drives me crazy is the glut of parameters that are involved. I understand that these objects are very complex, but I cannot help but break out into a sweat when I see "Let $G$ be a standard $(\alpha,V,\epsilon^2,k,\delta)$-pseudorandom widget...". Then I have to flip back in the paper or find another paper (which probably uses a different parameterization) and try to remember what all $\alpha,V,\epsilon,k$ and $\delta$ all meant.

It takes me quite a while to acquire a feeling for "good" parameter settings versus "bad" parameter settings, versus "natural" settings versus "easy" settings.

There's probably no magic bullet for this issue - but I was wondering if other people had some method of managing the "parameter explosion" so that it's easier to retain in memory for a longer period of time?


  • 2
    $\begingroup$ I am bad at pseudorandomness and I cannot talk about specific definitions. But I feel that “remember parameters” is a wrong goal. Parameters exist for some reasons, and it is more about understanding what must be parametrized, and that should naturally come out of your understanding of the notion to be defined. So basically your question is equivalent to “How do you understand a definition?” But then I do not think there is any silver bullet. $\endgroup$ – Tsuyoshi Ito Feb 3 '12 at 14:02
  • 1
    $\begingroup$ Having a personal wiki to keep track of definitions never hurts. This question might help. $\endgroup$ – Artem Kaznatcheev Feb 3 '12 at 15:51
  • $\begingroup$ HY its part of the curse of CS. try think like this. for example the defn of a simple turing machine written out actually involves quite a bit of parameters. but turing himself probably never wrote out the defn like that. therefore arguably many papers descr subjs in an unnec over-notated way. try to find author that explains the same concepts without so much notation/syntax. also textbooks tend to be great at simplifying complex papers better in hindsight if you can find that topic in the book. see the psychological concept of "chunking" $\endgroup$ – vzn Feb 3 '12 at 17:00
  • 2
    $\begingroup$ An addendum to my previous comment: see How do you get a “Physical Intuition” for results in TCS? The answers to that question (except for mine :( ) contain good, concrete advices about how to understand papers. $\endgroup$ – Tsuyoshi Ito Feb 3 '12 at 22:54
  • 1
    $\begingroup$ haha figured someone would ding me on that. imho turing has a lot of words where he could have symbols and a lot of symbols where he could have words. by the way another way of look at "too many symbols" is like "too twisty code" measured by something called "cognitive overhead" and its being seriously studied in some comp sci-cross-psychology research, for anyone interested try googling it. and there are recommendations on how to reduce "cognitive overhead" in coding that are directly applicable to paper syntax etc $\endgroup$ – vzn Feb 8 '12 at 5:28

My research practices, when I go into new research domain, cover combination of memory management, mnemonic, notes and other practices.

I have no one recipe, cause each is dependent on nature of given domain.

For some inspiration and sake of discussion, here is and example that came now into my mind:

Go through papers into a few iterations:

  • First, to get adjusted to domain and get first intuition about approaches and notations.
  • Next step is preparation to "clusterization" of papers.
    • prepare list of tags representing approaches, notations, features and other interesting me properties.
    • Before I start tagging I go through papers and evaluate my tags, correct appropriately.
    • Finally I tag papers
  • Then I process papers according to tags in groups: with similar notation, approach...
    • Thanks to working with papers sharing common properties, you can concentrate on differences.
    • I prepare notes in form of MindMap with FreeMind. To easily reach "what was what".
  • After processing papers grouped in sets with tags, try look at whole area with all groups to get boarder view.
  • Now is step of looking thought whole domain to see connections and differences. When I can't remember details, I check-out in my MindMap to remind, what was what.

Try to avoid overloading your memory, when it's unneeded, use: Mnemonics and Art Of Memory techniques. Use MindMapping.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.