Budapest trip

Last time I visited Budapest was the summer of 2002 during my IAESTE internship at Elcoteq in Pécs. This is a log of our trip during the summer of 2012.

Day 1

First time to fly Ryanair  – The flight was delayed a bit so we didn’t get to hear their jingle – Taxi booked online (20€) prior to departure – When we arrived there was an incident with an unattended bag in the parking lot but everything turned out to be OK – After checking in at the hotel, we went out for a walk to locate 0-24 hour shops nearby  and walked in Vaci utca and by the Danube.

Royal palace

Day 2

Breakfast at Cafe Gerbaud – Got the 72 hours transportation tickets (3850 Fiorints Per Person – FPP) for the metro, tram and buses – Started the walk: Vienna gate => Fisherman’s bastion => Royal palace => Tram to Gellert hill => Citadella => Liberty bridge => Small stop at the market => Raday utca for lunch and then Cafe Central for desert – In the evening we went for the standard walk on Vaci, Danube and Chain bridge – Later visited the Fashion street where they sell stove cakes (kurtoskalacs) and bread langos. As for the stove cakes, I liked the one with cinnamon more than the one with cocoa.

Fisherman's bastion

Day 3

These kurtoskalacs make for a great breakfast – Today’s tour started from Szent István Bazilika – Took the lift up to the Dome for panoramic view of the city – Then to Szabadság tér , a plaza with a pressure aware fountain and the Parliament – This was our first attempt to enter (tours sold out early since it was Sunday) – Margaret Island  – Walk to the end and took the bus back – Great Synagogue (largest in Europe, second largest in the world but largest in capacity) – English tour included in the ticket – “For sale” for goulash soup – Again the standard walk in Vaci, Danube and the Chain bridge.

Tram lines

Day 4

Woke up early – wait in the line to enter the Parliament (see my tip on 4sq and why you must book online first) – then off to market – tried fried langos  – the upper level of the market was pretty packed and seemed to me more of a touristy place than a traditional Hungarian spot – visited the renowned New York cafe.

New York cafe

Day 5

Szechenyi baths – Heroes square – Andrassy utca – Terror museum

Szechenyi baths

Day 6

Checked out and to the airport – This time the Ryanair jingle played upon landing

The Danube

Afterthoughts and observations

Even though there are not very famous museums to visit or monuments that stand out, I think Budapest’s beauty is in its location: the river, the banks and the bridges. We went to the terror museum since we found it to be different from others we visited in the past. The museum actually turned out to be quite atmospheric. Also a must do in Budapest is to visit one of the baths.

Prices are reversely proportional to the distance from Vaci utca from supermarkets prices to eating and change shops. For example prices in Raday utca, a nice street full of places to eat and drink, are much lower than in Vaci and of the same, if not better quality.

Tip for the supermarkets: Blue cap for sparkling and pink cap for non sparkling water.

A tip between 12-15% is included in the bill in around half the places we ate. In the others you can calculated it.

There were no big metro signs in the metro entrances, so you had to look for them. Also there are ticket inspectors at every metro we visited. I guess they had a big problem with missing revenue from free-riders and they resolved to this measure.

Budapest had enough tourists from all over the world, but not as many as I show for example in Barcelona last year.

More resources

I compiled a list of places I visited and some interesting tips in the foursquare Budapest trip list.

Also a small collection of photos can be found in the Budapest 2012 flickr set.

View Budapest in a larger map

Fitting a sigmoid curve in R

This is a short tutorial on how to fit data points that look like a sigmoid curve using the nls function in R. Let’s assume you have a vector of points you think they fit in a sigmoid curve like the ones in the figure below.

The general form of the logistic or sigmoid function is defined as:

y(x) = A + \frac{K-A}{(1+Qe^{-B(t-M)})^{1/\nu}}

Let’s assume a more simple form in which only three of the parameters K, B and M, are used. Those are the upper asymptote, growth rate and the time of maximum growth respectively.

y(x) = \frac{K}{1+e^{-B(t-M)}}

The following R code estimates the parameters, where y is a vector of data points:

Now the data points along with the sigmoid curve look like this, with a = 1.0395204, b = 0.1253769, and c = 29.1724838.

Translation of Echo State Networks in Greek

I think I am going with Ηχωικών due to Echo = Ηχώ. Thus Echo State Networks (ESN) = Δίκτυα Ηχωικών Καταστάσεων (ΔΗΚ) στα Ελληνικά. The idea basically came from thinking about the anechoic chamber = ανηχωικός θάλαμος.

Recursively count the number of files in a directory

I had a program that generated data files in several sub-directories inside a targetdir and wanted to watch the status of the process every now and then, by checking the count of the generated files. The command to do that in Linux is:

$ find targetdir -type f | wc -l

MsAriadne at Pac-Man vs Ghost Competition, CEC 2011

We got the third place with MsAriadne bot in the Ms Pac-Man vs Ghosts competition, organized by University of Essex and held during the 2011 Congress on Evolutionary Computation (CEC 2011) . The bot is part of George Matzoulas diploma thesis project.

Video of MsAriadne bot versus the top performing ghost team

Video of MsAriadne bot versus a random ghost team

Finding related work and keeping up to date

"Göttingen-SUB-old.books". Licensed under Public Domain via Wikimedia Commons.

One important task as a researcher, is to keep up with all the recent related work in your domain. The items below are what I do personally for following the state or the art. They involve both push notifications in the form of email alerts, and todos that I put on my calendar every two or three months.

1. I have subscribed to the RSS feeds of all well known publishers (Elsevier, IEEE, Springer etc.) for the journals I am interested in. Then using a RSS reader (I personally use Google Reader so that I have my feeds available and synced in all my machines and my mobile phone) in order to get all the articles in press. Every couple of months I spent a couple of hours to check the titles and abstracts. If the article seems interesting or is a related work I download it and read it further.

2. In my browser I have a bookmark folder for the websites of the research groups and researchers working in my area. Again every couple of months I do an “Open All in Tabs” action and browse through the tabs for newly added material. I put a recurrent reminder in my calendar app to check them every 3 months.

3. Through Google Alerts I have created a few queries that return newly added search engine results via email. Once per week new alerts arrive in my mailbox and I do a quick scan. For example a query could be “Echo State Network” in quotes in order to match the whole phrase. Also alerts can be added using Google Scholar, where one can save the searches as email alerts. I have a few of those too.

4. As in item 2, with research groups and researchers, one can have another bookmarks folder for conferences and workshops taking place every year. Some of them publish their proceedings online, so by visiting their websites two-three times a year you can step into the published work.

6. Last but not least, I have subscribed to a number of mailing lists in the areas of my interest. For example some of the lists I am subscribed to are:

  • rl-list (Reinforcement Learning)
  • ML-News (Machine Learning News)
  • reservoir-computing (Reservoir Computing)
  • cig (Computational Intelligence in Games)

among others. Sometimes authors advertise and provide links to their most recent publications, besides just using them for CFPs and job openings.

7. Currently I have started experimenting with social networking sites related to research like Mendeley and ResearchGATE. I’ll see how that goes.

What do you do?

Diploma Theses @ ISSEL on Computational Intelligence in Games

This is a video I made gathering clips from AI agents/bots/controllers or whatever you want to call them, developed by researchers, students and aficionados, mainly for competitions in the IEEE CIG conferences, with a couple of them being projects of diploma theses students at Intelligent Systems and Software Engineering Labgroup (ISSEL). The goal is to demonstrate existing test-beds to whoever is looking for developing autonomous agents as a diploma thesis project with ISSEL in the field of CIG.

The testbeds and related links


Car racing, car setup



Micromanagement, Small scale battle, Tech limited and Full game

Poker Texas Hold’em

Limit heads up, No limit heads up, Ring



Gameplay, learning, level generation


Keepaway (grid soccer environment)

Unreal Tournament

Video making, technical information

In case you are interested, clips were downloaded from the YouTube channels mentioned in the video, from attributed in the video websites, or with the help of screen capture programs in the following formats: flv, wmv, avi and swf. The swf video was converted to flv using a swf2flv converter and all of them to raw DV with the help of Kdenlive and Kino open source programs under Ubuntu Linux. For editing, rendering and uploading iMovie was used.


Along with Paris, Rome is one of my favorite cities so far. This was my second visit to Rome. The previous one was a short one, during an InterRail excursion a decade ago. Me and my wife decided to go there for our honey moon and this is just a journal of our time to the Eternal City. I’ve also tried to visit all the “Angels & Demons” sights and see them first-hand. So there will be an A&D post for sure in the near future. Our time of visit was end July – beginning of August 2010.

Day 1

Early in the morning caught our flight from SKG to FCO with Alitalia (they offered a drink and a snack to eat) – Got the Leonardo express (14 EPP) and in 30 minutes we reached the Roma Termini – Metro line B was out causing a little chaos in the bus terminals around – Check in – Visited the Castel Sant’Angelo and in particular: il Passeto (luckily since it was only open 10:30 to 11:30 am), walked through the castle and its museum halls (unfortunately no English translations available at the exhibits), terrace (nice view) – Short walk to the St. Peter’s plaza – Then headed east to Piazza del Popolo, Porta Del Popolo (Bernini) and Santa Maria del Popolo [1]: highlights there are the Cappela Chigi (Raphael’s & Bernini’s work) with the kneeling skeleton and the two Caravagios – Evening walk @ Via del Corso, Via Condotti, Piazza di Spagna and Scalinata, Fontana di Trevi and Piazza Barberini with Bernini’s Triton fountain.


Day 2

Colosseum (the 1.5 EPP booking online ticket fee saved us a lot of queue waiting time so I would recommend it) – Palatino: Museums, Stadio, Casa di Livia, Casa de Augusto, Roman Huts (not as exhiting as the Colosseum) – Roman Forum walk through the Via Sacra and Via dei Fori Imperialli (must have A LOT of imagination) – il Vittoriano: took the elevator to the top (kind of expensive with 7 EPP, but the view is really nice, personally I prefer it even from the St. Peter’s Dome since it is more in the middle of Rome than aside) – Piazza Venezia and Chiesa di St. Marco [2] (worth the “offerte” for lighting the golden mosaic) – Capitolium: Piazza and Musei Capitolini – Walk towards Theatre of Marcellus, Santa Maria Cosmedin with its Boca de la Verita (took a picture from the side, since it was not open at that time), the Broken bridge and Isola de Tiberina – At that time the clouds started gathering so we turned back to Circo Massimo where we took the metro back to the hotel after grabbing some take away food for dinner – The most tiresome day that put our legs to test, but at least it was worth it.


Day 3

Musei Vaticani (like in the Colosseum, the extra 4EPP for reserving the tickets online are worth it): just looked for the basic attractions in Pinacoteca, Museo pio Clementino, Gregoriano Egizio, Degli Arazi, Geographice, Stanze di Rafaello and Cappella Sistina – From piazza St. Pietro we entered the St. Peter’s Basilica [3], then the Dome, “Cuppola” in Italian, (elevator costs 7EPP) and finally then the Vatican Grotoes (for free) – The nice thing about going alone instead of being in a group is that you can take your time. Actually we stayed just in the Basilica for two hours – Back to the hotel… – In the evening a small trip to the Scalinata and Fontana di Trevi.

Burn to the end of time

Day 4

Plazza Barberini, Santa Maria della Vitoria [4] with its Theresa in Ecstasy sculpture by Bernini, Via Veneto, Santa Maria dei Concicione [5] and Crypto dei Cappucini (kind of kreepy, their motto: “What you are we used to be. What we are you will be” – In Via Veneto the coffee was expensive (5EPP) and we experienced some bad attitude from the waiters, unfitting for “want to stay famous” caffes – Walk in the park of Villa Borghese – Caught the nice view towards Piazza del Popolo – Sat in Caffe Rosati for ice coffee like proposed in AD (I bet Dan Brown has not visited Greece) for 7EPP – Headed towards Piazza de la Rotonda, the Pantheon, Piazza de la Minerva and the Elephantino obelisk, Santa Maria Dela Minerva [6] – Looked for the Caravagios @ St. Augustino [7] and St. Francesi [8] that were advertised in public spots all over Rome – Later Piazza Navona and Agnes in Agony [9] – Got some rest and went to eat – Grabbed an ice cream from the Old Bridge – Admire St. Peter’s piazza and Basilica at night until 11:00 pm, when they close it – Returned to the hotel after a quick walk at Bernini’s bridge in front of Castel Sant Angelo.

The Oculus

Day 5

Walking and shopping in Rome’s center – At afternoon, our attempt to locate Trastevere with no guide ends up in disaster, since we end up at Trastevere train station, which has nothing to do with the “cool” place in Rome – In between we saw the Pyramide and another view of Rome not so “historic”, but rather “urbanic” – Grabbed something to eat and returned to the hotel – After getting a good rest from the afternoon’s painful to the legs mistake, we relocated Trastevere, and went there on foot by the Tiber (pass four bridges heading south after Ponte Sant’Angelo and you will find it before Isola Tiberina on the west side of “Tiveris”) – @Trastevere: Santa Maria dei Trastevere [10], Piazza Trastevere, sat down to eat, searched for two Lonely Planet’s proposed gellateries but both were closed since it was kind of late – Went back again on foot, which we kind of regretted it since we felt kind of threatened in a situation. At least we got some nice night pictures with long exposure times.


Day 6

Checked out and left our baggage with the hotel – Got our small presents for family and friends mainly @ Via dei Rienzo – Made a final walk through Castel Sant’Angelo, Piazza Navona, Pantheon, St. Ignacio dei Loyola [11], Fontana di Trevi, St. Peter’s Basilica (Rome in general and St. Peter’s piazza more specifically by that time were packed with ROMA 2010 CIM attendees) – Finally, Hotel, Metro, Train to FCO, FCO to SKG, Thessaloniki, ate gyros and said home sweet home…



A small collection of my photos from Rome and the Vatican City.

Budget, Eating and Drinking

An expensive city (food, drink, museums). For our budget, we had to put some effort to think and search where to eat or drink, maintaining a good quality-to-money ratio. But I guess this is difficult for everything and everywhere nowadays. The coffee is not so cheap, as people often believe in Greece, especially in the caffes at the historic center. Kind of difficult though for the “average” tourist to see the attractions and have in his or her mind, where to eat “smart”: cheap enough and good enough. Personally, I prefer the pizzas and the freddo cappucino like made in Greece. Filled pasta was good (same for the lasagna) but the other kinds of pasta were too “al dente” for our tastes. Gellato was just great!!! Our personal favor was “Old Bridge” where we went three times. Fistacio there is just great. But beware, there are places where one can pay 5EPP (for example we spotted one such place at the Fontana di Trevi area) for an ice cream smaller than in places where you pay 1,5EPP for the same quantity.

In numbers

Photos taken: 1368
Churches (“Chiesas”) entered: 11 (including St Peter’s Basilica)


EPP: Euro(s) Per Person
AD: Angels & Demons

Recording an audio or video stream under Linux

Quick and dirty way to record an audio or video stream under Linux:

1. Find the URL of the stream:
2. Then: mplayer -dumpstream
3. To finish, kill the process
4. Open the file to listen or watch to the recorded stream

IEEE ICDM 2010 Contest

Just for fun, I participated in the IEEE ICDM 2010 ContestTraffic track, with a couple of R scripts, at first using linear regression and later neural networks. Mainly due to summer vacations limiting time available, the approach was nothing too fancy, ending up in the 17th place out of 101 active participants.

The task was to predict traffic in 10 road segments, 2 ways each, for 1000 60-minutes long windows between the 41st and the 50th minute, knowing only the first 30 minutes. Historical data were provided in the form of 100 10-hour windows (60000 rows) with 20 values per row, corresponding to the traffic observed in a minute of one of the 10 road segments x 2 ways.

My best result in the competition was obtained using the following procedure:

a) Preprocessing: Transform the training and test datasets, having corresponding to 10-minutes intervals rather than 1-minute intervals. Normalize all value to [0,1].

b) Modelling: Make the problem a supervised learning problem. I used 60 attributes, 20 for time t+1 to t+10, 20 for time t+11 to t+20, and 20 to t+21 to t+30, to predict one of the 20 traffic values at time t+41 to t+50. Thus 20 such datasets were created, one for each road segment and way.

c) Training: 20 Feed-Forward Neural Nets (FFNNs) were trained for each one of the above 20 datasets, and 20 more were trained the same way, using a reduced dataset with 15 attributes instead of 60. This was achieved by using ReliefF feature selection algorithm in WEKA and maintaining the top 15 attributes. Each one of the 40 FFNNs had its weights randomly initialized. The former 20 FFNNs had 15 hidden units, while the later 30. Decay rate was also used.

d) Predicting: Predictions were made for each one of the 20 target values using all 40 NN. The final prediction was the mean value of the 40 predictions.