The First Cry of Atom Today is the first day of the rest of my life.

Making storm cluster for development

When you develop big data processing platform such as Hadoop, Spark and Storm, you need to construct a cluster. You can create it with whichever virtual machines and real server. Personally it is hard to obtain real servers. Setting up networks and configurations are tough work. So you might use virtual machines, EC2 , VirtualBox. Today I’d like... Read more

Targets in 2015

This is my first post in 2015. There was a lot of things in last year and so will it this year. At the beginning of new year, I’d like to write down three targets I want to achieve. Deep Learning As this blog title is describing, I am very interested in artificial intelligence especially deep learning nowadays. I have kept learning deep learn... Read more

Training Conditional Random Field

This article is written as 17th entry of Qiita machine learning advent calendar. Conditional random field, CRF, is a kind of discriminative model for sequential data. This model is used widely for labeling natural language sequences such as “I have a pen”. There is a motivation to attach tags to this sequence. For example, “I have a pen” can be... Read more

Implement Random Feedback Neural Network

I’m very interested in neural network algorithm such as recurrent neural network, sparse auto encoder, restricted boltzmann machine and so on. Most of neural network learning algorithms are based on backpropagation. This algorithm was first developed in 1974 as the context of neural network. Backpropagation is a simple and efficient learning alg... Read more

How to launch WIP Spark cluster

In the process of development spark, it is not uncommon to investigate the performance of development version on EC2 cluster. However, this development version is of course work on process. So I thought it was difficult to launch with ec2 script included spark project. But today I found there is an easy way to do that with only one option. D... Read more