Machine Learning & Stats
13 talks from the 2022 edition.
-
`python-m5p` - M5 Prime regression trees in python, compliant with scikit-learn ▶ Recording 📝 Transcript
Sylvain Marié
Regression trees are powerful Machine Learning models capable of both flexibility in modeling as well as interpretability when the tree is not too deep. The M5 algorithm, introduced by Quinlan in 1992 to provide more...
-
Detecting drift: how to evaluate and explore data drift in machine learning systems ▶ Recording 📝 Transcript
Emeli Dral
When your ML model is in production, you might observe input data and prediction drift. In absence of ground truth, drift can serve as a proxy for the model performance. But how exactly to evaluate it? In this talk,...
-
Honey, I shrunk the target variable! Common pitfalls when transforming the target variable and how to exploit transformations.
Florian Wilhelm
Feature engineering takes up a huge part in the work-life of a data scientist. Sometimes this doesn't stop at features but also the target variable itself is transformed leading to all kinds of unexpected...
-
Inpsect and try to interpret your scikit-learn machine-learning models ▶ Recording 📝 Transcript
Guillaume Lemaitre
This tutorial presents the different inspection techniques currently available to inspect a machine-learning model developed with scikit-learn. In addition, we compare all these methods and pinpoint their limitations.
-
Introduction to Uplift Modeling ▶ Recording 📝 Transcript
Dr. Juan Orduz
In this talk we introduce uplift modelling, a method to estimate causal effects of a treatment, e.g. a marketing campaign, to effectively target customers that are most likely to respond to it. We describe the most...
-
Machine Learning Testing Ecosystem of Python ▶ Recording 📝 Transcript
Yunus Bulut
In this talk, I'll present the growing ecosystem of machine learning testing tools in Python. Machine learning validation and testing is an emerging concern in the MLOps domain and will become more so in the near...
-
Making MLOps uncool again ▶ Recording 📝 Transcript
David
Machine learning operations (MLOps) have gained attention among practitioners aiming to automate the development of Machine Learning models, attempting to mimic the impact of DevOps in software. However, MLOps...
-
My forecast is better than yours! What does that even mean? ▶ Recording 📝 Transcript
Illia Babounikau
Forecasting is one of the most popular applications of Machine Learning. In the last decades, it went from large numbers, few factors, and simple algorithms to small numbers, many factors, and complex ML models....
-
Predictive Maintenance and Anomaly Detection for Wind Energy ▶ Recording 📝 Transcript
Tobias Hoinka
This talk will describe predictive modeling applications in wind turbine maintenance, the challenges of anomaly detection and ways to move to more automatic diagnoses by modeling past documented defects.
-
Secure ML: Automated Security Best Practices in Machine Learning ▶ Recording 📝 Transcript
Alejandro Saucedo
As data science capabilities scale, the core concept of security becomes growingly critical. In this talk we will introduce the security challenges and solutions for data science practitioners relevant to each of the...
-
The secret sauce of data science management ▶ Recording 📝 Transcript
Shir Meir Lador
The question - “how to become a successful data scientist?” is often discussed in conferences like these. Today I would like to address a second level question. How to make the success scale? How to build a DS team...
-
Unsupervised shallow learning for fraud detection on marketplaces ▶ Recording 📝 Transcript
Andreu Mora
Combating fraud, scams and wrongdoings in large marketplaces and platforms that connect millions of individuals as sellers and shoppers poses a very exciting and also difficult problem. Adyen leverages massive...
-
You shall not share! ▶ Recording 📝 Transcript
Gönül Aycı
Online social network users frequently share personal information online. While each post is targeted to a certain audience, it is not always easy to judge what the privacy implications of shared content will be. To...