Introduction

I participated in PyCon 2016 JP! In this article, I will publish what the author, who is practicing Python acclaim, infiltrated (no, participated in) PyCon and heard about it. PyCon 2016 was held for two days, September 21st (Tuesday) and 22nd (Wednesday). I mainly participated in the program on the first day.

I'm ashamed to share my specs.

Work: Roppongi Web Service Company
Aimed career: Data scientist
Python usage: 10 months
Python level: 1 (assuming 5 is the highest)
The hardest Python coding ever: I built a DNN program with TensorFlow
Purpose of participation: Networking and knowing the level of the world

9:00 Admission-Opening Ceremony

Waseda University New Okubo Campus Building No. 63. Being away at all, I am confused but take a step toward admission.

PyCon has been established with the support of many sponsors, including Hunza, Monotaro, and gandi.net!

When you finish the reception, you will receive a name tag and a case to hang around your neck. Juice, coffee and breakfast were provided free of charge. Itareri Tsukuseri.

The opening ceremony was held in the large lecture room. Greetings from Chair Takanori Suzuki (in English).

Then there was a keynote speech from Jessica McKellar. McKellar is a San Francisco-based entrepreneur and software engineer who has also made significant contributions to PyCon in North America.

McKellar, who has been involved in open source development as a career background. He talked strongly about the impact and appeal of software and computer systems on people's lifestyles. Question time. So I asked ** "What do you think is the most elegant SYSTEM in the world?" **. In summary, "There are a lot of systems, and it's hard to answer in short time. But, in the first place, we should focus on the functionality and usability to the system, NOT whether it's elegant, or not." I got a reply like this. The level of speed of return and how to use Wit are different.

10:45 Listen to Blockchain for pythonists

Presenter: Mr. Abenben (Twitter: @abenben) Mr. Abenben has a career in system development for more than 20 years. I like Python and am an expert in financial system development. For beginners, he talked about an overview of blockchain and a Python case study.

Essence / Keywords

BlockChain originated from Paypal Mafia @ Silicon Valley
Adopted by 1370 companies worldwide
Unbundling (requires engineer know-how), rebundling, enhancement
Suddenly in fashion at Gartner's Hype Cycle 2016!
DLT distributed ledger technology
Ethereum (state of the dapps; provided by Microsoft)
Serpent
Hyper Ledger Project
R3 CEV
pybitcointools
It is predicted that ** Civic Tech (public) ** + ** Sustainability ** will come as "~ Tech" in the future.

13:00 Participation in a code explanation lecture for Python beginners

Click here for an overview → http://www.slideshare.net/tokibito/python-pyconjp2016 I focused on "os.path", "this", and "antigravity" in the library, and actually read the code inside.

14:30 Start chatops with Errbot

Presenter: Guillaume Binet chatops is a system that combines "Chat + DevOps", and he presented a Python program that combines a bot called Errbot.

Site: http://errbot.io/en/latest/ Core structure: user plugin -> git_hub storage plugins -> DB / Cloud Backends -> slack

16:00 Participation in invited lecture

Presenter: Mr. Seiya Tokui Preferred Networks researcher Tokui is leading the development of the deep learning framework Chainer.

The theme this time is 「how do we calculate gradient when there is "stochastic unit (a neuron with sampling)" ?」 The theme is to implement a model that incorporates a stochastic unit that propagates indefinitely into a unit in a neural network. This is a field that Mr. Tokui is currently working on. For the stochastic unit, "Gaussian Unit" and "Bernouli Unit" were proposed, and performances were given on how to implement each. Once implemented, it can be applied to generative models such as the Various Auto Encoder.

Mathematically, the goal is to calculate the expected value of the possible values of the unit from the probability distribution. (Because it is the representative value of the unit)

Until now, the "Likelihood-ratio (LR) method" has been the mainstream, but in addition, a method called "Reparameterization trick" has appeared.

The photo shows an example of Chainer implementation of a model that incorporates the Stochastic Unit into VAE.

17:30 Lightning Talk Five to six Talkers took the stage. At first I thought that a hard theme would be announced, but in reality it was a place for talks full of humor. The photo is the presentation of the first Taiwanese student. It seems that the karaoke scoring system was implemented in Python. It was with his own karaoke demonstration w

18:00 Ending Ceremony, Exchange Party

The theme of this year's PyCon is "Everyone is different, everyone is good" We were able to share the importance of developing and interacting with respect for Diversity with everyone in the venue.

at the end

I'm embarrassingly Elementary in terms of Python skills. Therefore, I hesitated to participate in such an authoritative place. However, I was able to actually participate and talk with players from many perspectives, and I realized that I was hesitant. When writing code, it looks like a battle with one person, but it was an event that made me realize once again that various inputs that cannot be seen with code grammar alone can be made through interaction. I'm glad I participated!

[PYTHON] [Report] Participated in PyCon 2016 JP