PyData Eindhoven Meetup - The first of many

After the very successful PyData Eindhoven event hosted at ASML and Greenhouse (https://pydata.org/eindhoven2019/), we wanted to start a regular monthly PyData Meetup in the Eindhoven region. The first PyData Eindhoven Meetup will be hosted by VersionBay (www.versionbay.com) at the High Tech Campus in Eindhoven.

There were many presenters who wanted to present at PyData but unfortunately, there were not enough speaker slots. We will start the PyData Event by reaching out to the people who submitted talks and did not get a chance to share their stories.

Agenda:
17:45 - Doors open
18:00 - Detecting scammers using AI, for real
18:30 - Dinner/Drinks
19:15 - A frequentist approach to prediction uncertainty
19:45 - Ensemble Modelling and Transfer Learning using TensorFlow 2.0
20:15 - Drinks and networking

----
Title: Detecting scammers using AI, for real
Description: HousingAnywhere is an accommodation marketplace with listings in 400+ cities around the world where advertisers publish hundreds of listings everyday. Keeping the platform safe and scammers-free is a challenge that we tried and managed to solve using Machine Learning. The one-year-long "Penelope project" resulted in a very large set of learnings both in the Data Science and Data Engineering fields that required us to redesign our Data Architecture and also to create a solid and shared Data Culture. From a technical perspective, we ended up building a complex ensemble model focusing on different aspects of a listing, and trained taking into account the evolution of the advertisers' behaviours during the years. In the subsequent iterations we also employed NLP to increase the performances leveraging the information retrieved from textual descriptions. This talk will try to give an honest review of the process, sharing some insights also on the best practices used to deploy Machine Learning models in production and to build and scale a Data Science Team in a fast growing company using Agile.

Presenter: Massimo Belloni - https://www.linkedin.com/in/massibelloni/
Company: HousingAnywhere - https://housinganywhere.com

Title: A frequentist approach to prediction uncertainty
Description:
Uncertainty for individual predictions becomes more and more important in decision making but is not always taken into account. Bayesian methods quickly come to mind when dealing with uncertainty, but there are frequentist methods to achieve (roughly) the same result. These frequentist approaches are often more easy to implement and computationally cheaper. This talk will explain how you can calculate different types of uncertainty (aleatoric and epistemic) for regression problems using quantile regression and Monte Carlo dropout and how to evaluate the non-continuous probability distributions. Also, a similar Bayesian model is implemented to compare with the frequentist approach. The concepts in this talk are accompanied by an implementation using PyTorch and PyMC3 (Code published on GitHub).

Presenter: Yu Ri Tan - https://www.linkedin.com/in/yuri-tan
Company: Xomnia - https://www.xomnia.com

Title: Ensemble Modelling and Transfer Learning using TensorFlow 2.0
Description:
Bringing Deep Learning algorithms to manufacturing and semi-conductor industry is a challenge. This is due to the steep learning process for newbies in Deep Learning. In this talk, I will illustrate how a newbie can adapt the best available Deep Learning models for problems at hand with ease using transfer learning and ensemble modelling.
Presenter:Anjan P. Gantapara - https://www.linkedin.com/in/anjangantapara/
Company: ASML

---
Attendee profile:

• People who want to learn about Python (Scientific Computing)
• Interested in seeing how others use Python
• Interested in being part of the Python Community in Eindhoven
• No Recruiters
• No Sales
• Only people who speak Python