PyData Tel Aviv Meetup: Lightning Talks 4
974 views · Published 6 September 2018 · 42:40 · Indexed 10 November 2026
Channel: PyData · 2018 · Science & Technology
PyData Tel Aviv Lightning Talks #4 23 August 2018 Sponsored and Hosted by Barrel Protocol https://www.meetup.com/PyData-Tel-Aviv/ Beware of data leakage / Yonatan Hadar - Data leakage is a very common pitfall that many data science projects fall into that may appear in many different forms, some of which are incredibly hard to catch. The effect of data leakage on the actual performance of production models and on the total effort spent on data modeling could be huge. In this talk I will cover some best practices of discovering and avoiding data leakage with some real world examples from my experience. Introducing the Hypernetwork: A Neural Network that Generates Weights for Another Neural Network / Lior Deutsch - Neural networks (NNs) are known to have many different optimal weight vectors. Where do these weight vectors lie in the "weight space"? Are they isolated from each other? This talk will discuss how to train a hypernetwork, which is a type of NN whose goal is to generate many different weight vectors for another NN. The hypernet can be used as a tool to probe the weight space, and it also has the practical applications of creating large ensembles of NNs, and improving robustness against adversarial examples. Multi-label classification - Methods & techniques / Alex Levin - In this lecture we will cover a various method and techniques to solve multi-label classification problems Detecting Obstacles: on the road and in your DL project / Noa Raindel - Advanced driver-assistance systems (ADAS) alert the driver in case of any danger. An example for such danger is an obstacle on the road (such as another vehicle, pedestrian, traffic sign, etc.). The aim of my project was to quickly detect such obstacles. One approach is semantic segmentation, which requires many computations and is too slow to perform in real time. Thus, a solution was found in implementing and improving upon StixelNet’s architecture. In this short talk I will overview the challenge in question, the solution I developed, and some personal insights I gained while working on a first deep learning project. www.pydata.org PyData is an educational program of NumFOCUS, a 501(c)3 non-profit organization in the United States. PyData provides a forum for the international community of users and developers of data analysis tools to share ideas and learn from each other. The global PyData network promotes discussion of best practices, new approaches, and emerging technologies for data management, processing, analytics, and visualization. PyData communities approach data science using many languages, including (but not limited to) Python, Julia, and R. Find a PyData chapter near you: meetup.com/pro/pydata 00:00 Welcome! 00:10 Help us add time stamps or captions to this video! See the description for details. Want to help add timestamps to our YouTube videos to help with discoverability? Find out more here: https://github.com/numfocus/YouTubeVideoTimestamps
More from this channel
-
31:23
Kostas Perifanos - Authorship Attribution Using Python
-
31:08
Prash Majmudar - Measuring the Digital Economy Using Big Data
-
36:26
Ian Ozsvald - The High Performance Python Landscape
-
37:48
Mike Muller - Shared Memory Parallelism with Python
-
35:01
Martin Goodson - Most Winning A/B Test Results are Illusory
-
30:19
Mark Basham - Python for High Throughput Science
-
1:13:42
David Cournapeau & Simon Jagoe - Narrowing the Gap Between Array Computing and Databases
-
31:38
Emlyn Clay - Phython, Pharmaceuticals & Discovery