PyData London 2016
2016-05-06..2016-05-08
London, United Kingdom
[PyData London 2016, PyData London 2017, PyData London 2018, PyData London 2019, PyData London 2022, PyData London 2024, PyData London 2025, PyData London 2026]
| title | author | length |
|---|---|---|
| Data Driven Diagnosis of Alzheimer's Disease | Frank Kelly, Giles Weaver | 37m |
A citizen science pipeline scrapes the Alzheimer’s Society forums with BeautifulSoup and produces user- and post-level dataframes (about 3,600 threads and… | ||
| Modelling a text corpus using Deep Boltzmann Machines in python | Ricardo Pio Monti | 32m |
Deep Boltzmann machines implemented in Python using bumpy model 10,000 abstracts as word‑count visibles with a softmax likelihood, pretrained as RBMs then… | ||
| Python vs Orangutan | Dirk Gorissen | 33m |
Six animals need locating across 2,000 km² so a drone-mounted receiver system is built that uses a software-defined radio feeding an Odroid running Python… | ||
| Deep Learning for QSAR | Rich Lewis | 40m |
Claims that deep learning is state of the art for QSAR and cites Kaggle Merck and Tox21 wins and then shows a reproducible pipeline: data pulled with… | ||
| Building a Pong playing AI in just 1 hour (plus 4 days training...) | Daniel K Slater | 50m |
Claim: build an AI to master Pong Pong in 1 hour (plus 4 days training...) using concrete methods: Q-learning, deep nets and Tensorflow. Implements tiny games… | ||
| Irregular time series and how to whip them | Aileen Nielsen | 30m |
Irregular series get concrete answers: use Lomb-Scargle periodogram (no imputation) to recover periods like RR Lyrae and the 11‑year sunspot cycle, prefer… | ||
| KEYNOTE: Scaling Out PyData | Travis E. Oliphant | 43m |
Claims that Python now scales both up and out with concrete tools and numbers: numba compiles Python to native code with GPU targets and reports up to 2000x… | ||
| A Gentle Introduction to Neural Networks and making your own with Python | Tariq Rashid | 56m |
Clear, concrete outcomes: a 3-layer network implemented as a simple Python/Numpy pipeline and shared as a Python 3 notebook on GitHub and… | ||
| Python flying at 40,000 feet | Marko Vasiljevski, Raffaele Rainone | 42m |
Shows a working pipeline that ingests black‑box HDF data, segments flights and detects the end‑of‑takeoff by modelling fuel‑flow and EGT around a 20‑second… | ||
| Estimating Residential Land Prices in the UK | Philippe Bracke | 41m |
Builds a merged dataset linking the LR Price Paid Dataset to Ordnance Survey MasterMap footprints and LR freehold parcel maps, using QGIS and Python to extract… | ||
| Probablistic Programming Data Science with PyMC3 | Thomas Wiecki | 39m |
Builds and runs Bayesian models in PyMC3 using a Student t likelihood and Hamiltonian Monte Carlo NUTS HMC to draw 5,000 samples, producing a posterior… | ||
| DyND: Enabling complex analytics across the language barrier | Irwin Zaid | 39m |
A C++ library that actually implements a C++ array container with a full type system and lightweight Python bindings is presented, supporting variable-length… | ||
| Making Recommendations without Data | Ruby Childs, Nick Sorros | 28m |
Recommendation system for Six Tribes bootstraps an interest space from Facebook posts and likes, iPhone music and photo/GPS location rather than historical… | ||
| Building Data Pipelines in Python | Marco Bonzanini | 38m |
Argues that reproducible, production-ready pipelines need concrete practices: unit tests and end-to-end tests early, Python workflow managers (demonstrates… | ||
| Estimating stock price correlations using Wikipedia | Delia Rusu | 31m |
Claims that Wikipedia-derived text vectors and graph measures can proxy future stock return correlations when price history is missing, tested on FTSE 100… | ||
| Real time association mining in large social networks | Ben Chamberlain | 31m |
Compresses the full Twitter graph — 240 gigabytes and 700 million accounts with 20 billion edges — into a few gigabytes by computing minhash signatures, builds… | ||
| Iterables and Iterators: Going Loopy With Python | Steve Holden | 44m |
Covers Python 2.2→3 iteration protocol with concrete code: getitem fallback vs __iter__/__next__, live IPython notebook and GitHub repo show a Stars class, a… | ||
| The CV: A Data Scientist's View | Rui Miguel Forte | 49m |
Claims that CV parsing is a rich, practical domain for data science; the team extracts text with Apache Tika but Tika yields messy, column‑and-table errors so… | ||
| KEYNOTE: Laser ranging in a new dimension | Andreas Freise | 47m |
LIGO achieves a clear detection of a gravitational wave from two merging black holes about 30 solar masses each, producing a waveform that matches General… | ||
| Lessons from 6 months of using Luigi in production | Peter Owlett | 42m |
Luigi runs dependency DAGs for ~500 jobs feeding a warehouse that supports 2000+ reports and 400+ BI users; Cron was replaced by EC2-based Luigi workers, with… | ||
| The NetworkL python package | Moreno Bonaventura | 38m |
NetworkL is a Python package that implements the Ramalingam&Reps algorithm and a sparse distance-matrix plus biconnected decomposition to shrink stored… | ||
| Cat modelling with python | John Gill | 38m |
A concrete account of what changed: legacy supercomputers (256MB, 3GB disk) get replaced by cheap Intel boxes and Raspberry Pi, compute cost plummets… | ||
| Classifying train and car journeys using telematics data | Annabelle Rolland | 41m |
A working classifier distinguishes train versus car trips from MyDrive Solutions telematics by engineering features like percentage of points snapped to roads… | ||
| 10 things I learned about writing data pipelines in Python and Spark. | Ali Zaidi | 39m |
Built a production Python + Spark pipeline that collates ~200 data points and computes ~300 metrics for ~80M account filings from ~11M private companies, uses… | ||
| Building a recommendation engine with Python and Neo4j | Mark Needham | 01h 38m |
PyData London 2016
In this session Mark will show how to build a meetup.com recommendation engine using Neo4j and Python.
Our solution will be a hybrid which… | ||
| Lightning Talks and Closing Address | Ian McKenna | 50m |
Video lists concrete tools and outcomes: a pip-installable envy-stripout that installs a git filter to strip notebook outputs (works to keep repos clean but… | ||
| The Duct Tape of Heroes Bayesian statistics | Vincent D Warmerdam | 42m |
Many concrete examples and code replace theory with results: a disease-test calculation that compares one versus multiple tests, a censored-exponential… | ||
| Using Support Vector Machines in Scikit Learn to discover genetic aetiologies | Tim Vivian Griffiths | 45m |
Support vector machines implemented with scikit-learn + Pandas + Joblib process features from single SNPs up to functional gene‑sets and fail to improve… | ||
| How to become a Data Scientist in 6 months | Tetiana Ivanova | 56m |
Claims a transition from project manager to hired data scientist in about six months and shows the concrete path: Coursera (Andrew Ng) coursework, joining… | ||
| Interactive Visualization in Jupyter with Bqplot and Interactive Widgets | Sylvain Corlay | 41m |
Shows concrete demos: installs from conda, live sliders wired to ipywidgets, plots made with Bqplot where every mark and scale is a synchronized widget, and a… | ||
| Julia for Data Analysis features, interfaces and future directions | Simon Byrne | 43m |
Claims Julia reaches C-like speed without leaving the language, shows JIT timing where a first run takes ~0.05s and later runs are orders of magnitude faster… | ||
| Twinkle twinkle little star, how I wonder what you are | Sandra Greiss | 36m |
Uses SDSS spectra rather than vague examples and shows actual pipeline choices: image BRISK descriptors and a non‑linear SVM trained on 8,500 DA and 16,000… | ||
| PySpark in Practice | Ronert Obst, Dat Tran | 43m |
Clear, actionable takeaways: configuration recipes (enable compression and speculative execution, reduce Python concurrency to avoid OOMs, use Anaconda on… | ||
| A B Testing: Harder than just a color change | Or Weizman | 44m |
Yelp’s Transaction Platform runs three experiments that together lift revenue 15 percent. A green-vs-orange button yields a 5.26% relative increase using a… | ||
| Spherical Voronoi Diagrams in Python | Nikolai Nowaczyk | 36m |
An implementation extends scipy to compute spherical Voronoi diagrams via numpy arrays and a 3D Delaunay/convex‑hull approach that finds circumcenters and… | ||
| Mining smartphone sensor data with python | Neal Lathia | 34m |
Collects smartphone accelerometer CSVs with an open-sourced Android app, extracts windows (5–20s), computes statistical/time‑series/signal features including a… | ||
| Building Data Pipelines in Python | Marco Bonzanini | 38m |
Argues that the hard decisions are concrete: use Luigi to define task dependencies and targets, package pipeline steps as reusable Python modules, and run unit… | ||
| Gotta catch'em all recognizing sloppy work in crowdsourcing tasks | Maciej Gryka | 29m |
Claim: a machine learning classifier reliably flags lazy crowdsourced QA so fewer repeats and cheaper checks. Uses captured mouse and keystroke traces plus… | ||
| Survival Analysis in Python and R | Linda Uruchurtu | 40m |
Uses Python Lifelines and R libraries survival and KMsurv, pulls author timelines from arXiv and enriches with genderize.io and affiliation parsing, fits… | ||
| To The Web and Beyond | Kyran Dale | 34m |
Claim: the most practical dataviz stack is Python for cleaning and serving and JavaScript for interactive rendering. Concrete tools and steps include using… | ||
| Challenges of analysing the wheat genome | Katie Barr | 39m |
Claim: wheat is triploid, five times more DNA than a human, and up to 80% repeats, so standard pipelines hit memory and logic limits. Alignment files balloon… | ||
| Python and Johnny Cash | James Powell | 43m |
Claims quick C‑Python hacks: AST-literals, resurrecting the print statement, lambda‑fusion. Demo builds an LD audit shim (≈10 lines of C glue) that embeds a… | ||
| Bayesianism and Survival Analysis | Jake Coltman, Jacob Goodwin | 01h 24m |
Models advertising path-to-conversion with Weibull distribution survival fits, handles heavy censoring, and estimates median lifetimes via MCMC using PyMC2… | ||
| Data Driven Diagnosis of Alzheimer's Disease | Frank Kelly, Giles Weaver | 37m |
Scraped Alzheimer Society UK forums yielding 3600 threads and 78,000 posts, cleaned with BeautifulSoup and aggregated by user; features include… | ||
| Robot detection in IT environments | Eszter Windhager Pokal | 36m |
PyData London 2016
In this talk, I will present the robot detection module of a machine learning-driven user behavior analytic tool. New methodologies were… | ||
| Statistically Solving Sneezes and Sniffles Step by Step | Ian Ozsvald, Giles Weaver | 39m |
Project records a year of one-person, timestamped symptom logs with GPS and backs them up to Segment IO, plus external London Air, weather, MyFitnessPal and… | ||
| Lies damned lies and statistics in Python | Peadar Coyle | 01h 10m |
Compares logistic regression implementations in Statsmodels scikit-learn and PyMC3 on the UCI Adult dataset, turning income into a >50k binary target (24%… | ||
| Word Embeddings for fun and profit in Gensim | Lev Konstantinovskiy | 34m |
Uses Gensim to load Google News Word2Vec (300‑dim) and trains on 2,000 movie plots (170k words); builds document features by averaging word vectors and with… | ||
| Hierarchical Bayesian Modelling with PyMC3 and PySTAN | Jonathan Sedar | 40m |
Compares PyMC3 vs PySTAN on a real dataset of ~2500 cars and the Volkswagen case, implementing hierarchical linear regressions with manufacturer and… | ||
| Working with Fashion Models | Eddie Bell | 44m |
A convolutional model trained as a multi-task multi-image CNN runs on multiple GPUs using Chainer, trained on about 4 million images (60M in-house) with ~60M… | ||
| Assessing the quality of a clustering | Christian Hennig | 47m |
Compares concrete clusterings (k‑means, Gaussian mixtures, PAM, hierarchical, DBSCAN, spectral) on datasets like Old Faithful and tetra‑gunilla bee genotypes… | ||
| Finding needles in haystacks with Deep Neural Networks | Calvin Giles | 36m |
Claims a learned manifold pulls duplicate product images close and different products far using a multi‑task classifier as a base, then metric learning with… | ||
| Bokeh for Data Applications and Visualization | Bryan Van de Ven | 01h 31m |
Bokeh is an interactive browser visualization library that demos a working stack: high‑level charts, the lower‑level plotting API, widgets, linked brushing via… | ||
| Introduction to Deep Learning & Natural Language Processing | Raghotham Sripadraj, Nischal HP | 01h 30m |
Repository is cloned and notebooks run; hands‑on includes MLP coded from scratch and a Keras Sequential MLP on a Kaggle auto dataset (input dim 93, dense 64 →… | ||
| Pandas from the inside | Stephen Simmons | 01h 02m |
Shows how DataFrames are implemented: data lives in numpy-backed block manager chunks, not as boxed Python objects, so single-column access is cheap —… | ||
| Python and Julia: the best of friends? | Malcolm Sherrington | 01h 00m |
Argument: Julia gains speed from staged compilation, type stability and multiple dispatch and Python gains access to Julia’s fast non‑vectorized loops via… | ||
| Deploying Jupyter Notebooks for students and researchers | Min Ragan Kelley, Thomas Kluyver | 01h 36m |
Shows concrete steps to run JupyterHub on one machine, then switch to DockerSpawner and GitHub OAuth, using conda or pip installations, configurable-http… | ||
| Flying Python at 40,000 feet | Marko Vasiljevski, Raffaele Rainone | 42m |
Real black‑box multivariate time series get cleaned, aligned from HDF quick‑access dumps and segmented into taxi/takeoff/climb/cruise/descent; the concrete… | ||
| Using Python with Hadoop to create production ready big data applications | Ulrich Zink | 44m |
Shows concrete pipelines: PySpark + MLlib ALS recommender on MovieLens deployed with spark-submit, backed by Spark as the data store and cached RDDs for… | ||
| Customising nbconvert how to turn Jupyter notebooks into anythi | Thomas Kluyver, Min Ragan Kelley | 33m |
nbconvert converts Jupyter notebook JSON into HTML LaTeX PDF and reST and is extensible via preprocessors exporters and postprocessors. Examples include a… | ||
| bandicoot: a toolbox to analyze mobile phone metadata | Luc Rocher | 36m |
Bandicoot is a free open-source Python toolbox that computes 160+ behavioral indicators from cellphone CDRs or CSV logs, plus reporting variables and… | ||
| Detecting novel anomalies in Twitter | Delia Rusu, Mattia Boni Sforza | 42m |
Anomaly detection flags finance-relevant Twitter spikes by modeling tweet counts per time window with Poisson regression with time‑of‑day day‑of‑week and… | ||
| Deep learning tutorial - advanced techniques | Geoffrey French | 01h 37m |
PyData London 2016
Some of the more advanced deep learning to help you get the best out of it in a practical setting. The main focus is on computer vision and… | ||
| Probablistic Programming Data Science with PyMC3 | Thomas Wiecki | 39m |
Overview of concrete claims and results: PyMC3 with Theano implements Bayesian models and runs MCMC (NUTS) and variational inference (ADVI) to automate… | ||
| Holy D@t*! How to Deal with Imperfect, Unclean Datasets | Katharine Jarmul | 31m |
Specific, actionable tactics appear: use pandas backfill/ffill/fill for time series gaps and watch for hidden nulls like repeated nines; detect duplicate… | ||
| The IceCube data pipeline from the South Pole to publication | Jakob van Santen | 38m |
A working pipeline moves ~1 TB of raw IceCube data per day from the South Pole, filters most of it on-site to fit a ~100 GB/day satellite link, and produces… | ||
| Statistically Speculating on the Source of Sneezes and Sniffles | Ian Ozsvald | 37m |
One year of self‑reported antihistamine usage plus sneeze counts and GPS traces feed a homemade iPhone logging app and HDF5/pandas pipeline; external features… | ||
| Assurance Scoring Using Machine Learning and Analytics to Reduce Risk in the Public Sector | Matt Thomas, Natalia Angarita-Jaimes | 33m |
A Python assurance scoring pipeline claims to identify and auto-route the majority of low-risk public sector cases so scarce investigator time focuses on… | ||
| nothing matches | ||