the Red Button

There is a question that I am obsessed with for more than a year since it appeared on my mind. Every person I’v asked gave me the “yes” answer. What would be yours?

Q:

You are in a white room with a red button. And there is a black suitcase with 1 million dollars in it.

You hear a sound from the speaker: ” If you press the red button, you will forget the last 10 seconds we will give you the million dollar.”

Would you press that button?

It seems like I can’t press that button no matter what. My conscious that includes the last 10 seconds will be erased. So I feel like the I, as a person who existed for some time including the last 10 seconds, who thought about things like pressing the button will be dead, and some person (me 10 seconds ago) will continue to live.

I just can’t press the button.

Can you?

Erkekler İçin Arkadaşlık Siteleri Analizi

İnternette hangi tip erkekler daha çok ilgi görüyor? Arkadaşlık siteleri(sosyal network) arasında farklılıklar var mı? Hangi arkadaşlık sitesinde hangi tip erkeğe rağbet ediliyor? Bu sitelerdeki bayanlar erkeklerden ne bekliyor? Karakterlerine göre farklı arkadaşlık siteleri seçen erkeklerin şansı yükseliyor mu?

Bu soruları cevaplayabilmek için amatör olarak bir sosyal deney hazırladık. Bu deneyde İstanbul’da yaşayan 3 adet farklı erkek profili ve 5 tane arkadaşlık sitesi kullandık. Deneyden çıkan sonuç şaşırtıcıydı:

• Çoğu sosyal networkte bayanlar iyi bir meslek sahibi, iyi gelir sahibi izlenimi veren, “düzgün ve elit” olarak tabir edilen modern erkeği tercih ettiler.
• Eğlenceye ve gece hayatına düşkün, cinsel içerikli mesajlar atan üniversite öğrencisi genç attığı mesajlara beklenenden daha yüksek cevaplar alsa da bu mesajların çoğu olumsuzdu. Ancak test sonuçları olumlu/olumsuz ayrımı yapılmadan sergilendi.
• “Kro” olarak tabir edilebilecek aşırı romantik erkek ise sosyal networklerin çoğunda aldığı geri dönüş ve bu geri dönüşlerin olumluluğu bakımından diğer 2 erkek profiline yetişemedi.

Metodoloji: Her sosyal networkte Random Sampling (Rastgele Örneklem) yöntemi kullanılarak, İstanbul’da yaşayan 600 kişi belirledik ve her karakterin hesabından 200’er kişiye mesaj attık. Toplamda 3000 adet mesaj gönderildi ve geri dönüş sayıları kaydedildi. Testimizin amacı nedeniyle karakterlerin tipleri profillerine uygun seçildi. Ancak yine de, profilden bağımsız olarak karakterlerin çekicilik seviyeleri sonuçları etkilemiş olabilir.

Devamı:

Bu çalışmanın tamamına PDF olarak buradan ulaşabilirsiniz.

Mahout Review by iletken

- This is a post I copied from iletken’s blog -

Taste is an open-source recommendation library in which you can create a basic recommender
easily.

In  this post, we will  try  to describe  some  key  features of Mahout’s Taste  implementation,  and  also discuss  some problems  associated with  it. Because Taste  has  standard  textbook  algorithms, we will  not  get  into  accuracy  details.  Therefore  our  main focus will be on performance. As a recommender system provider, we are also struggling  with scalability issues and we are investigating up to what extent a Hadoop implementation can help.  Therefore, we went  ahead  and  tested Mahout’s  latest  release  in  order  to  answer  thisquestion:

Does Taste in Mahout somehow solve scalability issues in recommender systems?

Sadly,  the  answer  was  not a  Yes.  Taste  is  a  great  API  which  is  very  easy  to  develop  with,  however  even  with  the  Hadoop’s  scalability  features,  Taste  is  very  limited.  Taste’s  original implementation is not scalable and this implementation does not fit into Hadoop’s map-reduce  logic  naturally.
After  providing  a  detailed  review  of  Mahout-Taste,  we  will  be  describing  some  of  the
methodologies  and  technologies  we  rely  on  to  overcome  our  performance  and  resource
management  problems with  iletken. We  will  be  recommending  some  basic  implementation
tips.

- Download the complete review in PDF format -

Summary
Pros:

• Mahout  manages  to  scale  training  session  with  slope  one  method  and  hadoop implementation
• Mahout is Open Source

Cons:

• Does not scale as expected
• Mahout scalability is achieved by only using a slope one method.
• Standard Slope one  recommender  is not very accurate  (Netflix  test: Success  -3%  (0.98RMSE))  compared  to  other  algorithms(not  included  in  Mahout)  such  as  Matrix Factorization(Netflix test: Success: 8.4% (0.87 RMSE))
• Inefficient implementation
• High memory & resource consumption
• Only Collaborative filtering
• Only standart algorithms

Edit: Some of our readers told me that mahout has improved. Please pay attention to what they are saying:

haltux:

“ Many criticizes are not true anymore.

The complete netflix dataset can be processed by the mahout slope one implementation with few hundreds megabytes.
Other algorithms are now available, included distributed ones.”

Goodbye

Goodbye my friend, Goodbye my hero.

Just moonwalk

Deniz OKTAR hakkında 25 ipucu

1- çok üşengeçim…

Scalability in Web Applications

When I met with a scientist from Last.fm on RecSys08, first thing he asked me about our iletken project was that: “how did you solve scalability issues? “ So, I have decided to talk about scalability and caching issues according to my experiences I had on my Internship at Turkcell Teknoloji. I will also try to explain a spesific product: Oracle Coherence, which is a clustered cache solution (or data grid). In the future, I will add some test results to this post. The main reason I am explaining Oracle Coherence is that I had the chance to work on it and they have very good examples/illustrations of data topoplogies which explains the different approaches of distrubted caches. There are many alternatives to Coherence but I do not have enough experience with them.

All the images below belong to Oracle.

If you are designing a customer oriented web application, you might want to consider the days when your users will exponentially increase. The mythical question is that will your application scale easily? Meaning that would your load capacity/performance of the application increase as you add more servers? The answer is not quite simple and don’t believe your IT department if they easily say “well… yeah sure. Why not? As we have more users, we will get faster or more machines “Scalability is actually very related to your initial design choices and it is hard to fix a non-scaleable system. So, scalability is not a hardware but a software design issue. A software architect must take scalability issues into consideration.

Strande prize :)

Check our website: iletken-project.com

Hi, check our new website for the iletken project: www.iletken-project.com

Success !

We have been chosen as one of the 5 finalists of “Strands $100,000 Call for Recommender Startups” with the project “iletken” which I have ben working on for 2 years. I will be presenting our project at the “ACM Recommender Systems 2008 Conference” at Switzerland on October 23-25.

It is a very joyful and encouraging development for us. The winner will get $100k funding but the most important part is to attend to the conference and meet with people who researches on recommendation systems. There were 25 competitors and this call was actually made for companies. So, as 4 collage students, we consider presenting at RecSys08 as a big success. Meanwhile, if we manage to get the $100k that would be just fine !

The other finalists are announced at RecSys08 website. I am excited to see that a there are two Netflix price competitor ( commendo (BigChaos) and Gravity RD)  in the list! Wow These two guys have the 2nd and 5th place.

If you are interested on what we are working on, this post includes some details: Modern Oil War of Web: Social Data War

Edit: I want to thank Turkcell Teknoloji for sponsoring some of our travel expenses for the conference.

iletken: murat deniz

9 sene sonra o adami bulmak ! Linux…

Mail gruplarına yazılan mailler konusunda dikkatli olmak lazım. Google da adımı aratınca hala karşıma çıkan pek rezil bir mail mevzusu var. Bundan tam 9 sene 6 ay önce, linux kurmaya çalışırken yaşadığım bir sorunla ilgili Bilkent Linux mail grubuna yazdığım birkaç mail seneler sonra ilginç bir süprize dönüştü. O zamanlar 14 yaşında bir orta okul öğrencisi olan ben, soru sormaya çalışırken komik tabirler kullanmış, cevap gelmeyince “yaaa noooluur cevap verin lütfeannn HELP HELP HELP” gibi devam mailleri göndermiştim.

(devami var)