I’ve been inspired by Yvain and ZM, so I wrote up my resolution, printed it, signed it, and taped it to the wall in front of my desk so I see it when I look up. All with a bit of ceremony of course. My full resolution is below. ZM inadvertantly provided some of the language. Feel free to copy and/or modify for your own resolution.
Also, the short time frame is due to my summer arrangements. On June 29, I fly to California to begin a 6 week internship. After I get a feel for how much time I can realistically apply to studying while there, I’ll write up a new resolution that takes those particular circumstances into account.
I, Matthew Simpson, realize that I am not a monkey brain, but am a timeless abstract optimization process to which this ape is but a horribly disfigured approximation. As such, I take it upon myself to improve this approximation.
First and foremost, I promise to continually remind myself that every minute and every dime is precious, and every minute and every dime that I don’t spend doing the best thing I can possibly be doing is a mark of sin upon my soul. Thus I resolve to spend every minute and every dime I have maximizing my utility function. I resolve to ask myself before every decision to spend money or time whether the chosen activity or good is utility maximizing.
In order to achieve this end, I promise to perform the following specific duties beginning Thursday, June 18, 2009 and ending Sunday, June 28, 2009:
Second, I promise to do mathematics for two hours a day, every day in order to prepare for the Iowa State Ph.D. Statistics program this fall. The television must be off during the math session and time spent talking on the phone does not count. Phone calls are not to be answered and text messages are not to be replied to unless there is a non social reason for doing so. The mathematics must be performed by working through the sections of and doing practice problems in the following books (in no particular order):
Probability: The Logic of Science, E.T. Jaynes
Probability and Statistical Inference 7e, Hogg and Tanis
A First Course in Real Analysis 2e, Protter and Morrey
Topology 2e, Munkres
Calculus 5e, Stewart
Elementary Linear Algebra 5e, Grossman
Elementary Differential Equations 6e, Edwards and Penney
Contemporary Abstract Algebra 6e, Gallian
The following exceptions apply:
Medical emergencies for myself, family, or friends, a car emergency, or other family emergency
Once during the period, a math session can be replaced with a two hour economics session using a suitable economics textbook that uses math extensively
Once during the period, a session can be skipped for any reason I deem fit
By attaching my signature to this document, I, Mathew Simpson, do solemnly swear on science, Bayes, and all that is rational to perform the above duties without exception, save those listed above.
Second, I promise to do mathematics for two hours a day, every day
But this is fishy, right? Because it’s easy to “do mathematics” for two hours every day without really learning anything. I’ve been thinking about the same kinds of problems (i.e. how to reliably learn mathematics) and one of my ideas is to use a formal proof checker. If you put yourself on a tough schedule that says something like “I will prove the first 10 theorems in PLoS by Wednesday”, then when Wednesday comes around you will understand those 10 theorems. The proof checker does not allow hand-waving; if it accepts your proof, you know you’ve achieved something. It also should permit moments of insight where you say “hey… this proof is clunky… what was Jaynes thinking? I can derive this result in 5 lines of HOL light!”
As long as I’m actually working through the texts, I’ll learn more than if I had not done the math at all, so it’s an improvement. Before my resolution, I had sat down to work through one of my texts exactly twice since I graduated and summer began. I’d been aware of my problem and wanted to do something about it for some time, but it seems my akrasia applies even to planning to do something about my akrasia.
This technique only works if you do what you commit to. Once you break your agreement, it stops working very well. You can work X amount, you cannot decide you will accomplish Y amount; what if it turns out one of the problems is much harder than you expected, or simply takes longer to work through, you will not get everything done, which will weaken the technique in the future.
I read Matt Simpson’s description of his intended process, and I have no idea what two “opposed techniques” you are talking about. Would you mind saying a bit more?
The idea is to apply the techniques to separate domains in order to see the relative strength of each. Not a synthesis, but more of a test. I’m pretty sure that specific duties will bound me, but I’m less sure about ZM’s technique. So I want to see what works.
If you want to study up on Bayesian stats, I’d recommend Bayesian Data Analysis, 2nd ed by Gelman et. al over Jaynes’s opus. There aren’t enough problem sets in PT:LOS, and the problems aren’t very relevant to the actual practice of Bayesian statistics.
Thanks. Right now, though, I’m constrained by the books I currently have. I just don’t have ~$50 to spend on an extra textbook. On the other hand, how does the first edition compare to the second? It’s at about $20 on amazon, which I may be able to do.
I’ve been inspired by Yvain and ZM, so I wrote up my resolution, printed it, signed it, and taped it to the wall in front of my desk so I see it when I look up. All with a bit of ceremony of course. My full resolution is below. ZM inadvertantly provided some of the language. Feel free to copy and/or modify for your own resolution.
Also, the short time frame is due to my summer arrangements. On June 29, I fly to California to begin a 6 week internship. After I get a feel for how much time I can realistically apply to studying while there, I’ll write up a new resolution that takes those particular circumstances into account.
But this is fishy, right? Because it’s easy to “do mathematics” for two hours every day without really learning anything. I’ve been thinking about the same kinds of problems (i.e. how to reliably learn mathematics) and one of my ideas is to use a formal proof checker. If you put yourself on a tough schedule that says something like “I will prove the first 10 theorems in PLoS by Wednesday”, then when Wednesday comes around you will understand those 10 theorems. The proof checker does not allow hand-waving; if it accepts your proof, you know you’ve achieved something. It also should permit moments of insight where you say “hey… this proof is clunky… what was Jaynes thinking? I can derive this result in 5 lines of HOL light!”
As long as I’m actually working through the texts, I’ll learn more than if I had not done the math at all, so it’s an improvement. Before my resolution, I had sat down to work through one of my texts exactly twice since I graduated and summer began. I’d been aware of my problem and wanted to do something about it for some time, but it seems my akrasia applies even to planning to do something about my akrasia.
This technique only works if you do what you commit to. Once you break your agreement, it stops working very well. You can work X amount, you cannot decide you will accomplish Y amount; what if it turns out one of the problems is much harder than you expected, or simply takes longer to work through, you will not get everything done, which will weaken the technique in the future.
How did it go?
That really seems like a bad combination. You are, it seems to me, trying to combine two opposed techniques without any real synthesis.
I think you’ve been voted down because your comment may be seen as unsubstantiated, as well as needlessly critical. Perhaps you’d care to elaborate?
I read Matt Simpson’s description of his intended process, and I have no idea what two “opposed techniques” you are talking about. Would you mind saying a bit more?
The idea is to apply the techniques to separate domains in order to see the relative strength of each. Not a synthesis, but more of a test. I’m pretty sure that specific duties will bound me, but I’m less sure about ZM’s technique. So I want to see what works.
If you want to study up on Bayesian stats, I’d recommend Bayesian Data Analysis, 2nd ed by Gelman et. al over Jaynes’s opus. There aren’t enough problem sets in PT:LOS, and the problems aren’t very relevant to the actual practice of Bayesian statistics.
Thanks. Right now, though, I’m constrained by the books I currently have. I just don’t have ~$50 to spend on an extra textbook. On the other hand, how does the first edition compare to the second? It’s at about $20 on amazon, which I may be able to do.
The Leatherby Library at Chapman University has a copy of the second edition (link). You’re going to be there in 8 days, right?
Wow, I didn’t even think to check their library and I’m the one who’s going to be there in 8 days. Thanks.