Alex Mikhalev blog

I am a bit fed up with MT, I looked at MT5 and as usual was seriously annoyed by MT backup behavior: I attempted to recover backup from mt 4 into clean mt 5 and hit usual wall with schema mismatch error. I believe recovery should work, no matter what. It should attempt to recover most of backed up data, even with error message.
And if I need to mess with database dumps and scripts, I am ready to move from MT.
Currently my choice is limited to:
WordPress – most of the people I read use it. Latex plugin works out of the box and MT import from text. Cons – I am sick of PHP code and if I need to hack it, I will be in seriously bad mood.
Jekill – can use git/hg for versions, I bet my computer is better suited for publishing latex with mixed html then server. Static pages generated. Cons – Ruby, I can’t say I know ruby, but I can hack the code.
Golbard – same as jekill but in python. I am pretty confident with python, but I will have to write movable type import.

As a big fan of automatic software deployment and cloud computing, today I tried chef following several tutorials one, two and three. But because I am lazy, I decided to try to use vagrant to setup chef server.

Read more…

Definition: We say that a vector of decision variables is Pareto optimal if there does not exist another such that for all and for at least one j. In words, this definition says that is Pareto optimal if there exists no feasible vector of decision variables which would decrease some criterion without causing a simultaneous increase in at least one other criterion. Unfortunately, this concept almost always gives not a single solution, but rather a set of solutions called the Pareto optimal set. The vectors corresponding to the solutions included in the Pareto optimal set are called nondominated. The plot of the objective functions whose nondominated vectors are in the Pareto optimal set is called the Pareto front.

In computer science it is common to measure algorithms performance by using O notation. Every book on algorithms development will start from this topic and although it is still good rule of thumb, the practical value of O-notation now becoming obsolete.

For example Mathworks dropped Flops since version 6, the command which was a good indicator of number of floating points operations in second. For another example see Is multiplication slower then addition, where Prof Daniel Lemire after perfoming basic tests concludes that “Hence, simple computational cost models (such as counting the number of multiplications) may not hold on modern superscalar processors.”

During my PhD thesis development, I compared different version Particle filter – stochastic algorithm, which performance depends on number of particles used with Hough Transform based algorithms, which performance depends on a size of the grid in accumulator. In this post, I will show how Pareto optimal points can be used as good visualisation techniques for algorithms comparison.

Read more…