Google maps thinks I can bike this 2,964 mile route in 10 days and 15 hours (after 1183 turns and no sleep). I'm less confident.

5/16/2010 Limits – talks about the limits on our bodies and minds. They claim that the human body is very conservative about its own limits, and will make you sore and exhausted long before you really come anywhere close to running out of energy. The energy gets saved just in case something truly terrible happens and the body needs to release some extra bursts. They describe the Race Across America, which sounds like an absolutely horrible experience.  People bike all the way across the country, more or less non-stop; sleep is “optional”.  The winner usually takes about 8-9 days, and bikes 22 hours each day. The radio show describes a rider from the Slovenian army who has won the race in five of the last seven years. Not surprisingly, he basically goes crazy after a day or two of pushing his body so hard, but this just leads to him having vivid hallucinations about being chased by enemy soldiers… in turn, these terrifying visions fill up his adrenaline and trick his body into giving him just a little bit more energy.

6/28/2010 Oops – this show covers various unlikely events and their even more unlikely consequences. One segment discusses the Berkeley Pit in Montana, a lake of toxic waste leftover from mining operations. The spot is of course a huge eyesore, but it was even worse in 1995 when a flock of over 300 snow geese landed in the pool thinking it would make a nice rest stop… they were sadly mistaken. Despite the poisonous nature of the pollutants in the pit, biologists have recently found a number of amazingly hardy microorganisms able to survive there.  One of them is even able to do a remarkably good job at transforming the waste into less harmful substances. The kicker: that particular microorganism has only been found one other place–the bowels of snow geese!

I’ve really enjoyed listening to these and other episodes of the show.  Sometimes the results from one segment of the show are used a bit too freely to make claims about later parts, but it is still quite thought provoking. The two hosts make a great pair, and give the show a nice style.  Highly recommended!

I just got back from a conference in Big Sky, Montana. The conference hotel was up in the mountains at a ski resort, and although they were starting to get their first snow of the season, the slopes weren’t quite ready yet. The scenery was quit breathtaking (especially for someone from New Jersey), but unfortunately I only got to walk around and take pictures during one of the cloudy days. The conference itself was quite interesting. It was much busier than any of the others I’ve been to. I met a number of new interesting people, and got to see a few other familiar faces which was nice.

I won’t bore you with the details, but you can all imagine it as collection of groundbreaking work exploring how virtualization can be used to make computers more efficient, improve reliability, and generally save the world from all sorts of evil.

It’s pretty exciting, and having finished that after 4 years puts me in a good position to finish my PhD well ahead of the average in my department…

I frequently have trouble communicating between systems which are separated by different private LANs. This often happens when you have your computer on a home wireless network which is separated from the rest of the internet via a router and firewall.  Fortunately, SSH tunnels can be used to link these machines together, provided you know which ports need to be interfaced and you have a common server which both machines can reach. Here is how.

Remote Access with SSH

Normally SSH is used to remotely log into servers in order to run commands or copy files. However, SSH can also be used to create “tunnels” between two servers. These tunnels are particularly useful for two reasons: they are encrypted and you can make network traffic going to one system magically appear at the end of the tunnel.

There are two basic kinds of SSH tunnels: remote and local. In all cases you create a tunnel that connects one port number on a remote server to a port on the local machine where you run the command.

Note: The commands here are primarily geared towards computers running linux or mac os x, although widows users can do the same if they install cygwin.

Remote Tunnels are used to forward all traffic from some remote port back to your own system. You would do this if you want to run a web server or some other service on your own machine, but the remote server can’t reach it because of your router. In effect, you are creating a listener on the remote machine which will accept the webpage requests, and forward them to your own computer.

You can create a remote tunnel with the following command:

ssh -R *:8080:localhost:80 user@remoteserver.com -N ## run this on home
## -R   = remote tunnel
## 8080 = port on remote machine that will accept traffic
## 80   = port on local machine to forward traffic to
## -N   = creates the ssh connection without opening an interactive shell
## -f   = (optional) use this to make the command create the tunnel
##         and return immediately.

This will make it so that all requests on the remoteserver.com to port 8080 are redirected to home port 80 (the default web server port). However, the remote tunnel only makes it so users on remoteserver.com itself are able to utilize the port forward. Thus localhost:8080 tunnels correctly from that server, but remoteserver.com:8080 will not go through the tunnel. To let a third user access the website, we need to also create a local tunnel.

Local Tunnels let you make a port on your own machine actually send traffic to a port on a remote machine. Often you need to combine a remote and a local tunnel if you want a user within a second private LAN to reach your machine.

ssh -L 8888:localhost:8080 remoteserver.com -N ## run this on work
## -L   = create a local tunnel
## 8888 = the port on the local machine for the tunnel entrance
## 8080 = the port on the remote machine

This will make it so that from the perspective of users on work, requests to localhost:8888 are forwarded to remoteserver.com:8080.  This in turn is connected to the original tunnel, creating the full tunnel back to home.

These kinds of tunnels can be very useful for connecting systems separated by routers and firewalls. Depending on the application you are trying to reach, you will need to use a different port number. Here is a good list of commonly used port numbers.

Of course there are some security implications here and your company IT people may not like these tunnels much. In effect, this lets you connect any two machines, even if they are supposed to be separated by secure VPNs. Of course, since only one port is being opened, and the commands generally aren’t left running for a long time, it shouldn’t be an issue.

Most of the information is pretty technical, so unless you are interested in doing some Bash shell scripting or need to make some regression plots in R, it may not be of much use.

BS and PhD Production

I was first surprised to see just how many PhDs are being awarded. Almost 1800 new PhDs were produced last year, and the growth compared to the 90s and early 2000s is pretty staggering. The report suggests that the numbers will be even higher in the coming year, although the rate of increase is expected to drop.

On the other hand, undergraduate CS enrollment is doing terribly right now. I’d blame this partly on bad stereotypes about computer science, but mainly on most undergraduates having no real logical basis for why they choose a major.  Sadly, I’m not sure if that will ever change.

PhD Employment

Next let’s look at where all of these PhDs are going after they graduate. Industry took 52% of the new graduates, while about 33% went to academic departments. Of the academics, more than three quarters went to PhD granting departments.

The trends for employment type are fairly steady, except immediately following the dot com bust around 2001. I suspect that the current financial woes in this country will again lead to an increase headed towards academia, although if the depression/recession is widespread enough it may impact academic funding as well.  Hopefully universities will do well since a lot more people are going to suddenly want to go back to school or stay in grad school due to limited job options.  On the other hand, my own department is not hiring anyone this year because of budget cuts, so it may take a while before the money appears to make this happen.

Of course CS is a wide field, and the subdivisions within it can act very differently, so we can dig down a step deeper to see the breakdown of new job types by area within CS. It is interesting to see here that OS/Networking (my own area) is by far the largest proportion of graduates. If you were to condense multiple groups into the three main areas in my department: Systems, AI, and Theory, then the systems group is by far the largest with 835 members, compared to 201 and 112 in AI and Theory (although an additional 627 are other fields like graphics or “unknown”). Note: these numbers are for grads who got jobs in North America only.

Although OS/Networking is the biggest group, about 75% of its members go to industry, compared to 55% of AI and 50% of Theory graduates. I’m not too surprised by this since Systems is of course a more practical field that can be more readily applied to industry goals.  However, it is interesting that on an absolute scale, there were more AI than OS grads going into academia (57 vs 51) even though there are almost twice as many OS grads in total. Hopefully that means that if I want to get a job in academia it will be easier since there are a lot of grad students in my field, but not many that are as interested in academic jobs.

What About the Money?

So, will getting a PhD make you rich? I’m not sure about that yet, but here are a few of the salary numbers that I managed to dig up. The Taulbee survey includes information on starting faculty salaries.  The tenure track median salary was $85,000.  Non-tenure track researchers can expect $66K and postdocs around $47K.  These are 9 month salaries, so the yearly numbers will be a bit higher (of course that means you have to write some grants to get someone to pay you for those other months).  I haven’t been able to get a good sense of how this compares to industry salaries.  This set of slides from CRA reports salaries in the $75 to $120K a year range for industrial research labs.  I’m not sure how industrial job salaries increase over time, but the mean salary for full professors in CS departments was $130K for 9 months.

So fiscally speaking, it sounds like you can get a pretty good job with a PhD (even ignoring non monetary benefits like increased autonomy and job stability). On the other hand you have to first survive 6 or more years of grad school without starving.  I did some comparisons of expected retirement savings comparing the scenario where I get an academic job at age 28 versus having immediately gotten a lower paying job out of college.  Unfortunately, the results were inconclusive — one site said that Professor Tim would run out of cash at age 99 and starve to death while another expected him to surpass his less educated counterpart at age 42 and go on to live off of investment interest for all of eternity.  I guess only time will tell.

Disclaimer: Most of these figures have been lifted from the survey website, and I encourage you to go there to see their full results and analysis. I have tweaked a few of the images to try to make them more readable, and created the last image using their text data.

Warning: In case the above sentence didn’t make this clear, this post is largely technical gibberish. If that’s not your thing, then you can move along and look at the pretty pictures from a hiking trip up Mt. Monadnock from back in September.

My feelings with CakePHP so far are mixed. As with many open source projects, the documentation for getting started is a little rough.  To help with that, here is the list of resources I’ve found most useful so far.

CakePHP Tutorials

IBM’s Cake Tutorial – 5 parts, including scaffolding and baking (topics not included in the Cake Book tutorial). Also has section on user authentication. In some ways the blog demo below is a better place to start since it gives you a more complete application quicker, but once you have the basic idea down this is full of good information, and I’ve only gone through part 1 so far…

The Cake Book Blog Tutorial – provides a good step by step guide for making a simple blog, but does not discuss the “bake” command or anything about personalizing the appearance of the site.

List of Cake Tutorials – links to lots of good information, although a lot of it is beyond my level right now. Hopefully I’ll get to the AJAX tutorials and do some nifty stuff.

General Documentation

The Cake Book – seems like it has a lot of useful information, including a guide to the MVC principles.  Some intro material is too basic, and other parts are too advanced, so it is a little hard to find the right pieces.

21 Things – some good tidbits not covered in the docs.

A Good Reference

Understanding Exposure by Bryan Peterson is an excellent way to quickly learn the basics. The book is full of beautiful photos with information on how to take them. He generally leaves out the important first step (move to a beautiful villa in southern France), but otherwise he covers what you need to know well. Before this I didn’t really understand how to start using the manual mode on my camera, and now I am almost brave enough to go ahead and shoot with it. (Update 12/31/08: I still do not use manual mode… oh well)

Online Photography Resources

• Jodie Coston – In depth articles on photography and technique
  
• Canon Enjoy series – Good Canon specific info with nice demonstrations of the effects of aperture and shutter speed
  
• Digital SLR Guide – lots of good tips
• Luminous Landscapes Tutorials – articles on a range of topics

Aperture

• Affects the depth of field (what is in focus)
• Using a higher f number results in a smaller aperture which gives a wider depth of field (more things in focus).
• Smaller f number -> wider aperture -> narrower depth of field (subject in focus with blurry background).
• Wide angle lenses are better for wider depth of field.
• Telephoto lenses are good for shots of a subject with a blurry background (narrow DOF).
• Online Depth of Field calculator

Shutter Speed

• Affects the blur of moving objects.
• High speed (1/1000) “freezes” the movement of object, but lets in less light.
• Lower speeds can lead to blurry, but brighter pictures.

ISO Settings

• Affects the picture quality/noise level.
• Controls the sensor’s sensitivity to light.
• A high value lets the sensor capture more light, but increases noise.

Exposure Compensation

• Adjusts how light whites and how dark blacks appear.
• Also may need to adjust this when a bright color fills a large part of the frame.

Canon “Creative” Modes

• P (Program): camera controls the aperture and shutter speed. Turning the dial changes the “program shift” which affects the depth of field.
• AV (Aperture Priority AE): You control aperture, camera does shutter speed. Press and hold the Aperture Preview button (on front near lens) to see what it will actually look like. This will decrease the light coming into the camera. Turning the dial will adjust the aperture size. Higher gives wider depth of field.
• TV(Shutter-speed Priority AE): You control shutter speed, camera handles aperture size.
• M (Manual everything): You control shutter speed and aperture size. Dial changes the shutter speed. Adjust the aperture settings by holding the AV+- button (back of camera, top right) and turning the dial.

Light Metering

• Controls the light in the scene
• Evaluative Metering: Camera controls light levels based on full scene
• Partial Metering: Camera uses area in center of viewfinder to determine lighting. Useful for backlit scenes with high contrast bright/dark areas.
• Center Weighted Metering: not clear if this is useful for anything.

(Auto) Focus Settings

• Determines what in the scene is in focus.
• One Shot AF: Pressing the button half-way makes the camera select which points in the scene should be in focus. The red dots for those focus points will briefly flash. Once the focus is locked, the camera can be moved (assuming the subject stays at same depth/position).
• AI Servo AF: Auto focuses on moving objects. After pressing halfway the camera will continuously refocus on target as it changes position.
• AI Focus: The camera decides whether to use one shot or servo focusing. Used for full auto mode.
• AF Point: You select which of the auto focus points the camera should use. Useful if camera is making something in the foreground in focus rather than the main subject.
• MF: Manual Focus mode. Useful if the camera is autofocusing on a fence instead of the rhinoceros behind it.

Bonus Neat Stuff

• Printable lens hoods

• Time lapse photos with a graphing calculator

I don’t have any sense of the quality of the different offerings–there are certainly some courses I’ve taken in the past which would be extremely boring to watch on video, so it’s not clear to me how useful many of these full lecture series courses would really be. On the other hand, some of the shorter “guest lecture” style talks might be more self contained and accessible.

A few that sounded interesting:

• Human Beings and the Machines of Sunshine – Rice
• Alan Turing: Codebreaker and Pioneer – MIT
• Death – Yale
• The Rise of The Creative Class – George Mason
• Cyber Humor: The End of Humor as We Know it? – Oxford
• Castiglione: The Art of Being a Renaisance Man – Syracuse

1. A good research notebook for informal notes and plans
2. A collaborative place to keep track of your results and general strategy
3. A way to keep track of your todos, meetings, and deadlines

Each of these has its own use, and none can completely cover for another. Here are my suggestions for each…

The Research Notebook

When I first started grad school I used a 99 cent notebook from the local drug store to keep all the notes about my project. It only survived for six months before the binding had fallen apart and I ran out of pages since I had been pulling them out to use for class notes as needed.

My current research notebook is a hardbound, 8.2 x 5.25, Moleskine with 240 sheets of graph paper. It costs about 15 times as much as the previous one, but it will definitely last a lot longer. Spending the extra couple bucks on it also has the psychological effect of making me automatically take better care of it. In fact, I’ve become rather attached to it.

I use the notebook to jot down quick notes or thoughts, as well as to plan out basic experiment and research plans. Of course it is also important to bring to meetings or seminars for taking down notes and comments. I do not record experimental results or exact implementation details in the notebook, as those are usually much better suited for an electronic format. Which leads us to…

The Wiki

I’ve become a huge fan of wikis. The point of a wiki is to be collaborative and make it trivial for both the original author and others to add comments and revisions. You can set up your own, but it’s even better to have a shared one for your lab. A lab wiki lets everyone keep track of what others are doing and gives a central, permanent place to store things such as experimental results and information useful to all lab members. Use it to quickly let others see your graphs and comment on them or to plan out your algorithms as a group.

Even if you use a shared wiki for your lab, you should keep a personal section where you can make your own notes. I find this particularly useful for recording basic code tricks, linux commands, installation notes, etc. The goal here is to make it as easy as possible for you to write down anything which might come in handy later. You think you’ll remember X in a year a month? No, you won’t. Trust me. So put it in your wiki and you can find it when you need it!

There are a ton of different wiki platforms to choose from. A simple one used by my lab which is convenient because it does not require any kind of special database is PmWiki. Even easier would be to use a wiki hosted by someone else such as pbwiki (as easy as making a peanut butter sandwich). Of course if you are doing super top secret research, you may want a more private place to keep your notes

If you don’t need something shared with others (or if it is just with a selected group) then you can also try something like backpack. It’s not exactly a wiki, but it provides an easy interface to create pages and add notes, images, and task lists to them. Pages can be shared with others or kept private.

The Calendar Task List

Now you have a place to take notes on your future plans and a spot to post your results, so what do you do next? Ah, well it seems you need a way to keep track of exactly that — your short term to do list. There are a lot of resources on the web for daily planning and keeping yourself organized. The GTD (Get Things Done) cult is gaining a lot of followers. I’ve read (most of) the book, and it is worth a look. The key idea is simple enough to implement on your own: have an organized place where you keep track of the list of “next steps” for each of your projects. These steps should be small ones, the idea being that if you always have a clear idea of your next manageable task, you are much more likely to do it than if your to do list for today just says “write Ph.D dissertation”.

There are a lot of online resources to help here. I’ve tried some larger life/project management solutions like Tracks and basecamp, but I think they are generally overkill for one person projects. Instead, I like the combination of an online calendar and to-do list. For the calendar, I use 30 Boxes since it has a very natural method for adding events (just type in “Meet advisor every thursday at 9am” and it will create the recurring appointment). I have been using Remember the Milk as a basic to-do list. It lets you easily create multiple lists and assign deadlines for each task to help you prioritize.

To make sure you actually use the calendar and list, you need to make them easy for you to see and edit. For this, I’ve been using netvibes to create a web portal which includes my calendar, to do lists, bookmarks, etc as embeddable widgets. Here’s what my page looks like:

From this one page I can then instantly see (and edit) my weekly calendar and my list of upcoming tasks. You can make your own page which includes calendars, bookmark lists, rss feeds, and all kinds of other content.

Conclusions

Sadly, even the most organized of us still have to spend most of our time actually doing the research which makes that organization worthwhile. There is also a certain danger of spending too much time getting organized and not enough time producing work to organize. In any case, there are undoubtedly ways that we can all further refine our work process so that we aren’t wasting time either figuring out what to do next or repeating what we’ve already done.

Resources

Wikis

• PmWiki – doesn’t need a database
  
• pbwiki – hosted by them

Calendars / To do lists

• 30 boxes – easy to use calendar
  
• Remember the Milk – nice to do list
  
• To do list roundup – more to do list options
  
• GDT Introduction – learn how to organize your life
  
• Tracks – full GTD system
• Basecamp – full project management solution
• Backpack – smaller scale self management with nice interface
• Netvibes – customizable start page made up of widgets

General Life Organization Resources

• 43 Folders
• Life Hack
  
• Lifehacker