PyData Amsterdam 2019
2019-05-10..2019-05-12
Amsterdam, Netherlands
[PyData Amsterdam 2016, PyData Amsterdam 2017, PyData Amsterdam 2018, PyData Amsterdam 2019, PyData Amsterdam 2024, PyData Amsterdam 2025]
| title | author | length |
|---|---|---|
| PyData Amsterdam 2019 | 03m | |
Conference overview explains PyData mission and format and then describes a healthcare ML project that estimates prediction uncertainty and applies… | ||
| Robots for Football | Wouter Kuijpers | 38m |
Tech United Eindhoven builds wheeled middle‑size league robots with onboard omni vision camera and a front Kinect sensor, runs CNNs offline for team‑colour ID… | ||
| Improving Machine Learning Workflow | Wilder Rodrigues | 23m |
As machine learning practitioners, we know how hard it can be to have a smooth process around training and serving production-ready models. Processing the… | ||
| The profession of solving (the wrong problem) | Vincent D Warmerdam | 35m |
Concrete outcomes and methods: a classroom regression gives an A+ but misses capacity limits; a recommender stack built with a database, cron/Airflow and A/A… | ||
| Sentiment Analysis on Sustainability Reporting | Susanne Groothuis | 22m |
Uses BERT base for transfer learning to classify sentence-level sentiment in corporate sustainability reports, creating a labeled corpus of 8000+… | ||
| Popping the Filter Bubble | Sanne Vrijenhoek | 19m |
A tool that measures diversity in news recommendations is built and released as open source to evaluate actual recommender outputs rather than debate the idea… | ||
| Build Facebook's Prophet in PyMC3 | Ritchie Vink | 28m |
Implements Facebook Prophet’s model inside PyMC3 reproducing the linear growth with 25 change points grid, vectorized dot‑product offsets and the seasonal… | ||
| The evolution of array computing in Python | Ralf Gommers | 30m |
Concrete bridge between NumPy and other array libraries: __array_function__ and __array_ufunc__ let NumPy functions dispatch to Dask, CuPy, PyTorch… | ||
| High on Cardinality, Low on Memory | Ondrej Kokes | 25m |
Four concrete tools and what they actually buy: plain bitsets for true/false arrays, Roaring using RLE to compress sparse bitmaps, Bloom filters are… | ||
| Uber's Synthetic Control | Nick Jones, Sam Barrows | 23m |
When a citywide change prevents A/B testing the team builds a counterfactual by weighting other cities and past data using Uber's Python package and a… | ||
| Online Machine Learning with Creme | Max Halford | 22m |
Claims that online learning removes retrain schedules and lets models learn per datum using an actual library called creme 0.01, showing an API with fit_one /… | ||
| Plant Factory: Sensor, Data, Machine Learning | Marianne Hoogeveen | 19m |
Claims an automated pipeline that places crops, schedules work, and optimizes environment using real sensor streams and vision‑based yield estimates; uses… | ||
| Having fun with Music and Keras | Marcel Raas | 31m |
Builds a simple NumPy synthesizer and an infinite synthetic dataset of notes, chords and drums, then trains a Keras model that embeds an FFT inside the… | ||
| Open-source City | Maarten Sukel, Swaan Dekkers | 30m |
Using Amsterdam’s open data portals and GitHub repositories, the talk presents concrete city projects that reuse publicly available datasets and code to… | ||
| Casual Notes about Causal Stuff | Lucas Javier Bernardi | 33m |
Practical claims about extracting causal effects from Booking.com data: a reproducible pipeline with propensity scores and two‑step reweighting / upsampling, a… | ||
| How to expand your NLP Solution to new languages | Lena Shakurova | 22m |
A CV phrase-extraction system uses cross-lingual embeddings (trained with word2vec on domain data and aligned via CCA) plus a BiLSTM+CRF sequence-labeler and… | ||
| Efficient Transfer Learning | Karlijn Zaanen, Satendra Kumar | 24m |
Booking.com’s MT pipeline uses OpenNMT-tf on TensorFlow with multi‑GPU training to scale from in‑house property descriptions to up to 96 million parallel… | ||
| Looking at Sound | Jurjen Feitsma, Wouter de Winter | 25m |
Raw audio becomes images by FFT into mel spectrograms (24 kHz, librosa) plus delta channels, fed to CNNs with inception‑style modules, frequency‑band splits… | ||
| Hacking Fitness | Jonatan Samoocha | 32m |
A probabilistic pipeline predicts amateur race speed from noisy watch GPS and heart‑rate data by fitting per-activity heart‑rate vs speed curves, bootstrapping… | ||
| A Walk Through the Isolation Forest | Jan van der Vegt | 24m |
Goes from the original Isolation Forest math to a concrete extension that Entropy Isolation Forests and the paper formulation lack: native support for… | ||
| 101 A/B Tests for personalized News | Ivo Everts, Matthijs van der Kroon | 23m |
Regiolab builds a central personalization platform for Omroep Gelderland using Primed.IO on Google Cloud Platform with Kubernetes, Spark, and Python. A/B… | ||
| SHAP and Beyond | Hidde Hovenkamp | 32m |
A gradient‑boosting model predicts seven‑day ICU readmission with AUC 0.82 using thousands of features and interpretable time‑series feature engineering; SHAP… | ||
| Low to High Resolution | Francesco Cardinale | 33m |
Keras implementation trains a roughly 120‑layer dense residual network on DIV2K (~800 images, 16‑patch batches, ~86 epochs on a p2.xlarge) and reports baseline… | ||
| Using Markov Chain approximation to identify financial fraud | Floriana Zefi | 25m |
Claims that sequential web data over six years yields predictive features and that a Markov approach helps spot anomalies in production are backed by concrete… | ||
| What's the uncertainty on your ML Prediction | Eva van Weel, Fabian Jansen | 20m |
A Python library errortools implements uncertainty estimates by approximating the loss as a multivariate parabola to get a Gaussian parameter covariance, then… | ||
| Building a Multi-Core Apply Function for Pandas | Diederik Greveling | 25m |
Claims a usable pandas extension implementing a multi‑core apply that keeps memory overhead low while speeding up slow apply calls. Explains and demos… | ||
| Remember What You Learn (Spaced Repetition Model) | Anni Sapountzi | 29m |
Claims that Half-Life Regression fits the Ebbinghaus forgetting curve to real learner logs and that Duolingo uses it to schedule reviews. Implements a simple… | ||
| Text analytics on annual reports | Anahita, Farokhi, Eltjona Qato | 23m |
A working Azure application ingests PDF annual reports, extracts text via pdftotree, runs NLP with comparisons between TextBlob NLTK and Google NLP then moves… | ||
| Annotating data the right way | Amit Beka | 29m |
Decide by concrete outcomes: uses Prodigy as the annotation tool, measures label quality with Kappa score and targets roughly 80% agreement, and prefers… | ||
| Speech Synthesis with Tacotron2 and PyTorch | Alexander CS Hendorf | 35m |
Implements Tacotron2 with PyTorch and trains on mel spectrograms to map text to audio samples, using an English public dataset of ~24 hours and a selfmade… | ||
| Genetic Regex | Robin Bakker, Ahmet Yigit Erdem | 13m |
An end-to-end system genetic programming to build regexes combines evolved patterns with a classifier that uses regex matches as binary features, trains and… | ||
| How to easily set up and version control your Machine Learning Pipelines | Stephanie Bracaloni, Sarah Diot-Girard | 01h 32m |
Promises concrete setup and shows it actually working: step‑by‑step project initialized with DVC and MLV-tools, data and intermediate files versioned via DVC… | ||
| Hands on - Build a Recommender system | Camille Couturier | 01h 31m |
Hands-on build of a movie recommendation engine that actually implements matrix factorization with SVD and a gradient-descent factorizer, uses the Surprise… | ||
| nothing matches | ||