New user interface

Hold down shift while pressing reload to make sure the page fully reloads if OptiMap appears broken.

The first thing you’ll notice is a complete makeover of the user interface. The functionality has grown considerably since I first posted OptiMap in 2007, and the controls were scattered randomly around. Now, they are organized into neat little drawers on the left, in the order you are likely to be using them.

There is some new functionality as well. The ‘Edit Route’ section gives you the chance to re-order the trip after it has been computed. Have to visit one client first? No problem, drag to the top. You can also delete stops from the itinerary here.

OptiMap 4 also has a more powerful TSP solver routine. The new solver kicks in when there are more than 15 stops, and produces better and more robust solutions than the old engine. It is guaranteed to perform better than before, and in some hard cases I’ve tested, it improved the solutions by as much as 6-7%. That means that if you spent 8 hours on the road before, you now get home half an hour earlier.

Support for Garmin GPS units is now added as well. Support is experimental, since I only had access to one Garmin device, which did not support routes. I’m interested to hear whether this works on your device if you have a Garmin GPS.

There are some bug fixes as well: When pasting data from Excel spreadsheets into the bulk add box, the cell delimiters were lost. This is fixed. The print function used to show half a page of unnecessary white space at the bottom, thus it sometimes would print blank pages at the end. This is fixed so you won’t be wasting paper needlessly any more.

As a bonus, you may now access the service from a shorter address: www.optimap.net (the old address will continue to work, and the page will remain identical)

I hope you enjoy the update! Please post questions, feedback, bugs and suggestions here. I’m sure I broke something during the update…

For a long time now, my mantra has been optimization. To make the most out of limited resources makes sense both economically and environmentally. OptiMap was an example of this. Driving too far is just stupid. Using OptiMap saves you time and money and it saves the environment too.

There are many other wasteful behaviors we need to address to optimize the usefulness (utility) of our planet’s finite resources. When I started to learn about embodied energy, I was shocked by how much energy is spent in the process of manufacturing the products we buy. An average $20 book has an indirect footprint of 20 kg CO2, mainly due to extraction of the raw materials needed and manufacture of the book itself. This is the equivalent of driving an average car more than 130 kilometers. Given the sheer number of things that most people own, these indirect emissions certainly add up.

How can we optimize this? First, we need to realize that in many cases, ownership of these items is not why we buy them. We want access to the features that the item can provide. You don’t buy the boardgame Axis & Allies because you want to be the proud owner of it. You buy it to play it. And it probably collects dust 364 days a year.

If someone has the item I need, and it collects dust most of the time, they wouldn’t mind lending it to me, right. And I would certainly return the favor some other day. But we need to survey the resources available. Then make that information searchable. So that when you need a power drill, you can find the one closest to you. SkyLib.org is my attempt at doing this. Yes, it’s beta and yes it lacks many features. Certainly rough around the edges. But you can start organizing your stuff and building your personal library today.

Some features that are not done yet but under development:

• Friends support, invites and import from other social networks
• Buying and selling things
• User ratings
• Wanted-functionality (I need x)
• Renting and deposits
• Custom location search (Find item closest to me)
• Facebook login
• Embedding products on third-party websites
• Intro-video

Feedback is hugely appreciated! Please head over to www.skylib.org and give it a try. I’m already sharing 120+ items with you as of this post.

Consider the TSP problem with N+1 points labeled 0, 1, …, N. We start at point 0. The distance from point i to point j is dist[i][j].

In dynamic programming, we seek to solve a problem by first solving smaller instances of the same problem. We start by looking at the really trivial size problem: What is the best order to visit just one of the destinations?

Assume we have N destinations. There are N size-1 problems, one for each of the N destinations. The best way to visit each is the shortest path from node 0 to node N. We store the best solution to each of these problems in a table:

best[subset][end] = dist[0][end]

Here, subset represents which destinations have been visited. More specifically, subset is a bitmask: an integer where bit i is 1 if destination i has been visited and 0 otherwise. Also note that the stored solution is the solution to the A-Z TSP problem, so it does not include the cost to return to the origin. The variable end is the destination that this A-Z problem ends at.

Now, lets expand to all the A-Z TSP problems of size one larger:

best[subset][end] = min(best[subset \ { end }][i] + dist[i][end])

Here, i is an intermediary node that is 1 in subset and not equal to the end variable. Before we move on to the next size problem, we have to fill in all the possible subsets of the current size. Finding all of those subsets of size s is what the function nextSetOf(s) in BpTspSolver.js does. How do we justify this expansion? It does not matter to the distance of the larger problem in what order the intermediary destinations of the 1-size smaller problem ending at i was visited in, so we are safe to only store and look at the optimal such sub-solution.

Finally, when we reach the desired size where all the destinations are visited, we only need to look up the best solution to the A-Z TSP problem in the table:

best[2^(N+1)-1][N]

(2^(N+1)-1 is simply the integer with all the first N+1 bits set to 1). If we want to return to the origin, we instead take the min over all i of

best[2^(N+1)-1][i] + dist[i][0]

This then becomes the best solution to the roundtrip problem. It is important to note that while this solution may look fast, there are 2^(N+1) different subsets involved (really 2^N, because 0 is always visited). Also, each subset solution needs to be stored, causing high memory usage. However, O(2^N) is still much better than the brute force O(N!) solution. 10! is 3 628 800 while 2^10 is only 1024. With the brute force solution, OptiMap could solve optimally up to 9 destinations. The dynamic programming solution now allows OptiMap to solve optimally up to 15 locations.

After being “almost ready” for way too long, the next version of OptiMap, based on Google Maps API version 3, is now launched. While I have tested most aspects of the application, there are most likely bugs, since the update touched almost all areas of the code. There are some improvements too, however:

• More robust address lookups (a lot of people experienced a “failed to geocode” error when entering many addresses. This was due to too many requests in a short period of time, and a queue-system has been added to avoid this from happening. The lookups may take slightly longer due to this.
• Progress indicator for directions lookups. Because version 3 of the Google Maps API only allows 10 waypoints in a single request (down from 25 in the previous version), this part is now a bit slower, so a progress indicator is needed.
• Tuning of the solver code based on faster browsers becoming more common. This should improve the quality of the solutions for cases with more than 15 locations.

Please help me solve any bugs that you may encounter by posting a comment on this site. Information that will be helpful when locating the bug includes:

• Browser (with version if possible)
• List of addresses or locations and a description of how to reproduce the bug
• The output that you see (error message, why you think the solution is wrong etc.)

I’ve created a map which shows the measured radiation values in Japan (note that this map is no longer online – the rest of this post is out of date). The data is scraped (credits go to Marian Steinbach) every 10 minutes (hit refresh to get the newest data). You can also click on each measurement location to see a chart of the measurements from that station over time.

Currently, there are about 200 measurement stations, but I’m having trouble finding the latitude and longitude of each measurement station. I’m sure someone who knows Japanese would have more luck…Anyway, I’m slowly working my way through the list of measurement stations, so more locations will be added continously. Any help with this would be much appreciated. Feedback and criticism is welcome, and should be added as comments to this page.

Update 2011-03-20: With the help of volunteer “hosoyamane”, a Japanese translation of the map is now available. It’s great to see volunteers pop up so fast!

Update 2011-03-20: A small spike in radiation is showing up in the stations in the Ibaraki prefecture around 10 am Japan local time this morning. However, the levels are still low (1000 nano-Gray is still 2000 times less than the average yearly dose of background radiation).

Update: The original radiation map I created is no longer maintained. Instead, users are redirected to a map created by the Institute of Information Design of Japan.

Let’s look at some numbers:

And what has the S&P done?

This is looking to me like a very unhealthy market. It’s like the high commodities prices of the first half of 2008 all over again. Only this time, the economy is probably a lot less resilient towards high prices. If the economy tanks again, what instruments remain to bail it out? The “wealthy” nations of the world are already running close to 0% interest rates, and are up to their ears in debt. Printing money, as has already been done, is a surefire way to boost commodities even higher. Peak oil, the climate crisis (drought brought forth some of the high grain prices seen above) and general resource depletion are all converging to shake the foundations of our debt-based consumerist society.

The Davos growthsters have only one plan: Returning to the good old years of growth by “stimulating” the economy by ever more illusional money.

I have another idea: It’s time to question that old growth gospel and stop borrowing money and natural resources from our future.

A 1-litre plastic (PET) bottle has an embodied energy of roughly 5.4 MJ. This is the same as leaving a 15 W lightbulb on for 100 hours (1 kilowatt-hour is 3.6 MJ, a 15 W CFL lightbulb gives light equivalent to a 75 W incandescent light bulb). Aluminium cans are about 1.7x worse than plastic bottles in this respect.

Recycling changes the picture a little. Recycled PET has an embodied energy about 20-25% lower than virgin PET. Other types of plastic can achieve reductions in embodied energy of up to 50% by recycling.

While recycling can reduce the impact of disposables (and keeps the plastic out of landfills and nature), the impact of recycled materials is still huge. Avoid disposables whenever possible!

Source: The ImpEE Project, The University of Cambridge, Recycling Plastic


0.98^40 = 0.45
0.97^40 = 0.30


What this means is that if you start saving at age 27 and want to retire at age 67, a 2% annual fee will shave 55% off your retirement account. A 3% annual fee will take out more than 70% of your savings.

This model is too simplistic, since most people will save a certain amount each year and add to their retirement savings. These subsequent amounts are not subject to the fee as many times as the amount you put down in the first year, e.g the amount you save the 2nd year is only subject to the fee 39 times, the 3rd year 38 times etc.

The annual returns you get is another complicating factor. Given higher annual returns, these annual fees should matter less, right? Not so, it actually doesn’t matter at all what the underlying annual return on the investment is – the fees will cut the same percentage of your total life savings, since


(1+p)^n / ((1+p)*(1-q))^n = 1 / (1-q)^n
where
p := annual base return on investment
q := annual fee


I even created a more advanced model which takes into account inflation, increases in salary, various return rates on the investment etc. You can copy it and play with it from here: Google Spreadsheets

The spreadsheet model shows that the 3% annual fee is not as bad as the initial calculations show, but still ends up eating half your savings under reasonable assumptions (income = 100000, tax rate = 33%, inflation = 2%, pay increase = 3%/year, spend = 50000 first year, then increasing 3%/year, return on savings = 7%)

Do managed mutual funds offer higher returns than passive index funds, which have much lower fees (typically 0.1% – 1%)? Various studies have tried to answer this. A survey of the existing literature is given in Performance Persistence for Mutual Funds: Academic Evidence. The various papers quoted in this article show different results. At best, there is very weak evidence of some skill among the top fund managers, but no study found that the funds generated excess returns after fees were subtracted. It also shows that past performance is a very weak predictor of future performance – after two years, the is virtually no correlation at all. So if a fund performed great until now, there’s no reason to believe it will continue much longer into the future.

Let’s end with a few quotes:

“A respect for evidence compels me to the hypothesis that most portfolio managers should go out of business. Even if this advice to drop dead is good advice, it obviously will not be eagerly followed. Few people will commit suicide without a push.”
Paul Samuelson, Economist, Nobel Laureate

“If “active” and “passive” management styles are defined in sensible ways, it must be the case that, (1) before costs, the return on the average actively managed dollar will equal the return on the average passively managed dollar and, (2) after costs, the return on the average actively managed dollar will be less than the return on the average passively managed dollar. These assertions will hold for any time period. Moreover, they depend only on the laws of addition, subtraction, multiplication and division. Nothing else is required. ”
William F. Sharpe, Professor of Finance, Nobel Laureate

Will you continue to let banks, traders and fund managers fleece you, to the tune of half your retirement savings, for a service which can be fulfilled just as well by a monkey throwing darts? Or will you change to passive index funds, and give these leeches what they deserve?

Here is an alternative, if you want some market exposure: In the US, the ticker SPY tracks the S&P 500 index at an annual cost of 0.1%. In Norway, Handelsbanken offers its XACT OBX fund which tracks the main index at Oslo Børs for a fee of 0.3% a year. These trade like normal stocks.

I’m not saying this is a good time to get involved in the markets, but there’s never a good time to pay too much in fees!

My personal experience with banks was increased somewhat over the last month, as I tried to wire transfer several thousand dollars from a bank account in Canada to one in Norway. The accounts are both owned by me. They are with large, respected entities in both countries. I have several receipts (thank God) that show that the account number (IBAN), SWIFT code, address, name etc. provided by me are correct.

The money did leave my bank account in Canada a few days after I requested the transfer. So far, so good. The problem is that that was in December 2009. After waiting for a week, I first contacted the sending bank in Canada. They gave me a nice looking “Payment Transmission Confirmation” showing the information I had entered and some other numbers. The transmission showed that the money had left my bank, then gone to an intermediary associated with my bank and had been sent on to Bank X (I know the name, but won’t post it here yet) in Great Britain, with instructions to send it on to my bank.

I gave my receipt to my bank in Norway, but they could not find any funds, not in my name, not to my account, not of that amount, not on that date, etc. Their intermediary is another Norwegian bank, so I tried calling them – no luck. They pointed me to DnB NOR, who is the intermediary they use. DnB NOR was also unable to find the money.

I requested more information from Canada. Another week went by, and all I got was this very detailed spreadsheet-looking SWIFT request from my contact in Canada. The first receipt had some numbers. This one had lots of numbers! I sent it on to my Norwegian bank, who didn’t quite understand it. While it certainly had numbers, it didn’t have the number they wanted (transaction reference number).

At this point, I directed a request for a formal investigation to the originating bank. It’s been a week since then. They are no doubt calling banks down a long chain of intermediaries now. What I know about the trace of my money is this: My Bank in Canada -> Intermediary in Canada -> Intermediary in Great Britain 1 -> Intermediary in Great Britain 2 -> unknown. After the unknown, the final part of the route should be: DnB NOR -> Intermediary in Norway -> My Bank in Norway.

If you think this looks a bit like how the internet (tried traceroute?) works, you’re absolutely right. The only difference is that this process is manual. Yes, you heard it right. What happens at each of these links, is that some peon in the sending bank faxes a document with instructions and numbers on it to the next bank in the path. This is done over a “secure” (physically separate from the normal telecommunications network) wire. (Thus the term wire transfer). Then some peon in the receiving bank will look at the instructions, type the info into their system, and if the destination is not yet reached, fax it on to some other bank. The woman I talked to at DnB NOR even suggested that at some point in the chain, they had failed to find a “path” (translated from Norwegian “sti”) towards the destination! The routing in SWIFT is completely ad-hoc you see.

Whenever you visit a major city (at least in Canada), you can bet that the tallest five buildings are banks (not counting the CN Tower of course). If you’ve ever wondered what the people in those 100-story glass palaces do, now you know: They are full of peons who form the human equivalent of CISCO routers in a giant human version of the internet. In 2010, I believe this meets the definition of Epic Fail. And yes, the banks have ways of making us pay for this largesse!

View the certificate

Looking forward to go back to Norway! I’m sufficiently alienated here in Canada that it’s good to return to a sane (relatively speaking) place once in a while. Hope to see some of you there!