The first cry of Atom

The people who are crazy enough to think thay can change the world, are the ones who do

about me

I’m a software engineer living in Tokyo, Japan.I have a passion for developing the artificial intelligence that is the dream of humankind for a long time.I long for the world where technology supports our imcompleteness which is burdened since human was born. So I dedicate myself to realize these ideal some day.

How to Do Pitch

Sometimes I have a chance to do pitch or presentation for private or for my work. Anyway I always feel that is the first time to do so because I cannot make advantage of past experiences. This is a challenging task.

This is the seminar I cannot help reviewing my presentation skill. This is about iOS8 and Swift.

written in Pitch, Presentation, Swift, iOS Read on →

Write Octopress Article on Two Machines

By accident, I got MacPro. This is my first desktop machine in my life. I usually use notebook or laptop machine because there were a lot of time to bring my machine outside. But now I found it is very confortable to use it, large size display, high spec for processing and sufficient amount of capacity.


written in Blog, MacPro Read on →

Build Hadoop on MacOSX

I am not a veteran Java developer. I don’t understand the detail of Java development environment such as build tool or some type of libraries. Today, I realised that I need to develop or understand architecture of Hadoop. Hadoop seems to use some kind of Java development tools. Maven, ProtocolBuffer, CMake and so on. So in this article I want to record my environment for building hadoop projects on MacOSX.

written in Hadoop, MacOSX Read on →

Living in Silicon Valley

I am not living in SF. My house is at Tokyo, Japan now. Moreover it’s the first time to visit Silicon Valley. I don’t know what I can understand from what I saw, what I felt and who I met. So it’s my review time for this trip and make use of this to know how I should live. There are 3 points.

written in Silicon Valley Read on →

Case Class for Swift

Last week, Apple new programming language Swift was released. From that time I keep considering Swift looks like Scala language. Scala has two sides as object-oriented-language and functional-programming-language. So there are many features you should learn from scala. One of the most powerful feature of scala is pattern matching. This feature in scala context can be applied to all type objects. It is called constructor pattern matching.

written in Case class, Scala, Swift Read on →

Welcome Open Source Movement in Hardware Company

Today is the first day of new era for Open Source movements.

All Our Patent Are Belong To You

Tesla Motor by Steve Jurvetsonflickr

Tesla is famous for its motor products. These opened patents includes EV technologies for the future. I think it will accelerate the speed of development of EV vehicles. So this decision plays a big role to progress the bright future of motor technologies.

Above all, as a software engineer, it is great thing to spread open source movement into hardware companies.

written in OSS, Open Source

Hack NHK

Recently, there are a lot of hackathon in Japan. This word “Hackathon” also have been familiar word to non programmers. People want to use this type of event in order to solve any problems they have at office, home and life.

NHK is Japan’s national public broadcasting organization. The name is abbreviation of “Nippon Housou Kyoukai”. This organization keeps focusing to improve the quality of news broadcasting in terms of equality, accuracy and deepness. In this weekend, Hakathon is applied to the field of broadcasting. I attended.

written in Hackathon, NHK Read on →

What Is the Essential Factor of Autoencoder?

The other day, I wrote neural network which implements backpropagation algorithm. Following this program I write denoised autoencoder program by inheriting previous neural network. Backpropagation algorithm is generally good performance in spite of the simplicity. With this code, I can be ranked in at the 266th(May 20th, 2014). So I think this implementation has no bugs. However when I use this program as autoencoder the same is not true. With autoencoder, you should reduce dimension of input vector in order to extract essential characteristics. These essential characteristics might be also reduced so it cannot reconstruct the same vector to input vector. In fact after over-completed this network, I can better performance in terms of the number of accurate answers.

This is code.

written in Deep Learning, Machine Learning Read on →