[PYTHON] Matters examined at the August 2017 study session

This is what I investigated at Nagoya.Swift + August Study Group --connpass.

From this time, we will also describe the matters investigated by other people.

Machine learning

I was looking at the source code of magenta and training Udacity.

magenta

The activity of project magenta, which generates music by machine learning, is still active.

The procedure for moving it is also published on Qiita.

The output result when I tried to move it with midi of Touhou Project (Imperishable Night) according to the above procedure was like this. Somehow it smells like red free paper.

When I noticed, I was surprised that I was able to output drums. I want to try it.

kokoro

When I was looking at magenta, there was a folder with the name magenta / kokoro, so I was trying to find out who it was.

When I referred to 2016-jenkins-world-jenkins_inside_google.pdf, the CI used inside Google Sounds like the name of the tool.

Udacity

I'm asked what I recommend for machine learning training, but I think Andrew Ng's Machine Learning | Coursera is good.

However, I think that I need to be quite enthusiastic and secure time (I was frustrated about 3 times), and I also wanted to do it with Python instead of Octave, so I recommend the Google course. ..

You can see the code of the person who is working on it in advance, so I think it is also popular.

However, in most cases, copying does not work. For example, tf.concat seems to be a little different from the past version.

  ix = tf.Variable(tf.truncated_normal([vocabulary_size, num_nodes], -0.1, 0.1))
  fx = tf.Variable(tf.truncated_normal([vocabulary_size, num_nodes], -0.1, 0.1))
  cx = tf.Variable(tf.truncated_normal([vocabulary_size, num_nodes], -0.1, 0.1))
  ox = tf.Variable(tf.truncated_normal([vocabulary_size, num_nodes], -0.1, 0.1))
  # Doesn't work
  # sx = tf.concat(1, [ix, fx, cx, ox])
  sx = tf.concat([ix, fx, cx, ox], 1)

The example I worked on this time is like this.

  1. LSTM The content is to learn using Wipedia data obtained from About the Test Data and to speak something English-like. It's almost random before learning.
ot zmwpnsdlhbt orh g uhn x q dzilnnjkheet rdu jaf zakdfeaf gka ez pa fc saj seo
 oiipenfkniemohxxiuugdbsmntyae njrzze ysptv g f ucwzjstprjetktnr rqu sgkqeeg wgn
rio gxiaagltsenb zp ihacvzhlruwniafer ribrignq nljsc vaucyttmzhienn u rwhq gu ii
n on mdpy mn eybllngudeekm lzcorecvhhgxeniyuxq  fkk ato  k yl sunksjcinafgjnak i
cybdd etizneliudnsrnoqam  hcemiqip ue lioffhe y ujq eyjpyxiebdvqvyn pczitcvssft

After learning, it will be like this. You can see English words.

voly stated beath or were mhyn spains consenty prime importa studied to duc comm
y partiather conspition in rishen alpharry in with the stage simpos fabouting th
way earl war scokiu as the film formbus painpents their flagned are sternincrati
ameritially coderish adomecord lyhas orgers spipal days and for log while minocr
 a side the rearory the air mager have days yorge as medi american outsom be cou

Tensorflow

The great thing about Tensorflow is that it has a lot of tutorials.

This is one push.

Building a Web Application development environment

The web application development environment was created in two ways.

Vagrant + VirtualBox

A web application development environment was built using Cent OS. It seems that the procedure was referred to the following video.

Docker Compose

I was building a web application (Nginx + Rails + MySQL) using Docker Compose. I also found the following steps in Qiita.

Creation of presentation materials

There was a person who created a presentation material at HackMD and created a commentary on MNIST of TensorFlow.

Here is the material that was created.

Next time preview

The next event will be held on Saturday, September 16th.

Recommended Posts

Matters examined at the August 2017 study session
Matters examined at the study session in May
Study at Zundoko
I tried studying on the WEB server side at an in-house Python study session
Read "Ordinary Linux Programming" at an in-house study session