Powered by Big Data: How more data improves Machine Learning.

Y-DATA Meetup #8
Powered by Big Data: How more data improves Machine Learning.
18.11.2019 18:00

Hosted by WeWork
Talks are in English

Intro:

In recent years, data has been accumulating at an unprecedented rate. This has been propelled by the growth of the internet, smartphones, online social networks, and the internet of things (IoT). These enormous piles of data have disrupted nearly every field, from agriculture to healthcare to politics. The amount of collected data achieved the volume required to start building deep E2E models that outperform heuristic- and feature-based models individually crafted by domain experts. Such models can provide value both independently and in concert with work by domain experts and can even offer new, previously unknown insights. In the following talks we will present several real world use cases for this process.

More info about Y-DATA is here: bit.ly/ydata-website
Previous meetups videos are here: bit.ly/youtube-ydata

Agenda:

18:00 - 18:30 Registration, Mingling, Snacks & Beer

18:30 - 19:15 Talk #1: Real-world examples of How Big Data enables ML - Aleksandr Dolgarev, CTO at Quantum

19:15 - 19:30 Break

19:30 - 20:15 Talk #2: The Science of networks: Big Data in action - Michael Fire (PhD), assistant professor at BGU

Talk Details:

Talk #1: Real-world examples of How Big Data enables ML:

1. Building the E2E model based on 10-year claims data for epilepsy patient diagnosis
2. Processing imagery data for tree cuttings detection

Abstract:
In-depth look at two use-cases for a powerful mix of ML with Big Data.

• The first is in the healthcare area in the US, where experts were able to use claims data for the past 5-10 years to successfully improve epileptic patients' classification as well as inferred new knowledge regarding this difficult and unpredictable disease.

• The second use case is clear-cut segmentation on satellite images, where, with the help of data from Sentinel-2, it became possible to automatically analyze the vast amounts of forested areas, which was not possible before.

Despite being very diverse examples, they share common problems such as data pipeline and data storage building, data privacy, interpretability results, model certification, etc. These cases perfectly demonstrate a common trend: more and more areas will benefit in the future from ever-growing data that allows relying not on rules or features but raw data and deep models.

Bio:
Aleksandr Dolgarev is CTO at Quantum. He leads data science projects for pharma, infrastructure, machinery market leaders.

Talk #2: The Science of networks: Big Data in action

Abstract:
This lecture presents our recent studies that use machine learning (ML) algorithms, combined with network science, to analyze a variety of datasets. Namely, we will demonstrate how big data and datasets, ML, and network science can be used to:

• Predict relationship connections in social networks
• Identify fake online profiles
• Estimate people’s lifespans
• Uncover an individual’s personal traits using network data
• Predict students’ exam scores
• Estimate if a movie passes the Bechdel test

Our goal is to identify significant patterns and create useful predictive models using state-of-the-art techniques in data science.

Bio:
Michael Fire is a Senior Lecturer (Assistant Professor) at the Software and Information Systems Engineering Department at Ben-Gurion University of the Negev (BGU). Michael’s main research interests lie in big data, machine learning, social network analysis, and security and privacy. Michael also has gained extensive hands-on experience as a data scientist working for several companies and organizations.