Difference between revisions of "Learning"

Revision as of 11:04, 12 November 2012 (view source)

Green26 (Talk | contribs)

← Older edit

Latest revision as of 14:16, 1 May 2016 (view source)

Green26 (Talk | contribs)

(Blanked the page)

(2 intermediate revisions by the same user not shown)

Line 1:

−

~~'''Learning''' [[User:Green26| (alec green)]]~~

−

~~----~~

−

~~Observations on maximizing learning productivity, where 'l-unit' (learning unit) is used as a compact representation for any equation, method, theorem, fact, etc. The form is as follows:~~

−

~~<pre>TITLE~~

−

~~. analogy~~

−

~~. claim~~

−

~~. justification~~

−

~~. counter~~

−

~~</pre>~~

−

~~----~~

−

~~[1] 'MAXIMIZING SHARPNESS'~~

−

If your goal is to climb to the top of a rock wall the fastest, you don't blindly adhere to a straight vertical path up (you might run out of footholds), nor do you just move horizontally sideways (you won't make any progress upwards). rather, you move vertically until your rate of advancement is slowed to a point that sidestepping the obstacle above and scaling a new vertical path would result in a faster net movement upward.

−

~~Likewise, switch subjects often (for me, 2-3hrs), attempting to reach some 'goal' over each interval (digest a book chapter or video lecture, implement something in code, complete a proof).~~

−

This keeps you alert, and avoids inherent decay in productivity due lack of 'pressure' or due to boredom with material. Additionally, this will force you to recall l-units in an 'unprimed state' (not 'recently' exposed to media explaining the given l-unit) a greater number of times. The ability to recall an l-unit in an unprimed state is what we often want as the outcome of learning, and is proportional to number of times you've attempted to recall that l-unit before.

−

~~However, if you jump around too much, you may suffer significant 'context switching', or undesirable overhead associated with refamiliarizing yourself with a given subject.~~

−

~~[2] 'NONLINEAR LEARNING'~~

−

~~When you chop down a tree, it often becomes useful to chop from different directions, otherwise, your axe/saw may get 'pinched' and not make much progress.~~

−

~~Likewise, don't constrain yourself to a linear learning path through a subject, in which you don't move ahead until you fully master the current material. Attack the subject from different directions.~~

−

How an l-unit works (operation), and what it's used for (utility) often don't become clear until you look further into the material, where the l-unit is compared and contrasted against similar l-units. Therefore, it can be useful to superficially memorize the formula/method for the time being, and move ahead with the intention of returning to earlier material--in other words, jumping forward and working back. I've inefficiently spent much time trying to understand pedantic details that turn out to be trivially easy to understand once I progress further into the material.

−

A balance must be made, however, so that you still have the discipline to spend the multiple hours often required to comprehend a concept. Otherwise, you may learn a lot of high level concepts with no ability to implement them. ('If you can't build it, you don't understand it.')

−

~~[3] 'CONSISTENT PEDAGOGICAL SOURCE'~~

−

..

−

~~Extract content from a pedagogical source (in-person teacher, textbook, video lectures) of consistent presentation/teaching style as long as possible.~~

−

..

−

In practice, humans are 'experts' (relative to other humans) in a limited number of domains, so one should analyze whether a new or current pedagogical source should be pursued when learning a new subject. Eg, learning computer architecture from an English teacher who has mentored you for five years is likely to be more inefficient than learning from an entirely new source. Additionally, despite the claim of this section, it may turn out that the maximum rate of learning occurs by only learning one subject from each source, so that in practice no source is ever used for more than one subject. An intuitive balance (defined by the individual) must be struck between using the same source or a new source in order to maximize rate of learning.

Difference between revisions of "Learning" - Rhea

Latest revision as of 14:16, 1 May 2016

Alumni Liaison

@@ Line 1: / Line 1: @@
-'''Learning''' [[User:Green26| (alec green)]]
-----
-Observations on maximizing learning productivity, where 'l-unit' (learning unit) is used as a compact representation for any equation, method, theorem, fact, etc. The form is as follows:
-<pre>TITLE
-. analogy
-. claim
-. justification
-. counter
-</pre>
-----
-[1] 'MAXIMIZING SHARPNESS'
-If your goal is to climb to the top of a rock wall the fastest, you don't blindly adhere to a straight vertical path up (you might run out of footholds), nor do you just move horizontally sideways (you won't make any progress upwards). rather, you move vertically until your rate of advancement is slowed to a point that sidestepping the obstacle above and scaling a new vertical path would result in a faster net movement upward.
-Likewise, switch subjects often (for me, 2-3hrs), attempting to reach some 'goal' over each interval (digest a book chapter or video lecture, implement something in code, complete a proof).
-This keeps you alert, and avoids inherent decay in productivity due lack of 'pressure' or due to boredom with material. Additionally, this will force you to recall l-units in an 'unprimed state' (not 'recently' exposed to media explaining the given l-unit) a greater number of times. The ability to recall an l-unit in an unprimed state is what we often want as the outcome of learning, and is proportional to number of times you've attempted to recall that l-unit before.
-However, if you jump around too much, you may suffer significant 'context switching', or undesirable overhead associated with refamiliarizing yourself with a given subject.
-[2] 'NONLINEAR LEARNING'
-When you chop down a tree, it often becomes useful to chop from different directions, otherwise, your axe/saw may get 'pinched' and not make much progress.
-Likewise, don't constrain yourself to a linear learning path through a subject, in which you don't move ahead until you fully master the current material. Attack the subject from different directions.
-How an l-unit works (operation), and what it's used for (utility) often don't become clear until you look further into the material, where the l-unit is compared and contrasted against similar l-units. Therefore, it can be useful to superficially memorize the formula/method for the time being, and move ahead with the intention of returning to earlier material--in other words, jumping forward and working back. I've inefficiently spent much time trying to understand pedantic details that turn out to be trivially easy to understand once I progress further into the material.
-A balance must be made, however, so that you still have the discipline to spend the multiple hours often required to comprehend a concept. Otherwise, you may learn a lot of high level concepts with no ability to implement them. ('If you can't build it, you don't understand it.')
-[3] 'CONSISTENT PEDAGOGICAL SOURCE'
-..
-Extract content from a pedagogical source (in-person teacher, textbook, video lectures) of consistent presentation/teaching style as long as possible.
-..
-In practice, humans are 'experts' (relative to other humans) in a limited number of domains, so one should analyze whether a new or current pedagogical source should be pursued when learning a new subject.  Eg, learning computer architecture from an English teacher who has mentored you for five years is likely to be more inefficient than learning from an entirely new source.  Additionally, despite the claim of this section, it may turn out that the maximum rate of learning occurs by only learning one subject from each source, so that in practice no source is ever used for more than one subject.  An intuitive balance (defined by the individual) must be struck between using the same source or a new source in order to maximize rate of learning.