PyData Chicago 2016
2016-08-26..2016-08-28
Chicago, United States
who it's for
data scientists, Python developers, machine learning engineers, data engineers, analytics engineers, open source maintainers, infrastructure engineers
topics
pandas and scikit-learn dtype bridging; scikit-learn pipelines deployment; distributed grid search with joblib; dask parallel computing in Python; luigi for data pipelines; PyMC3 Bayesian modeling; elastic net generalized linear models; Feather columnar format internals; TensorFlow beginner workflows; ImageNet feature extraction for EDA; Kalman filter forecasting; building recommendation engines with Python
| title | author | length |
|---|---|---|
| Embracing the Monolith in Small Teams | Leon Sasson | 41m |
Argument for a monolith focuses on specifics: small cross‑disciplinary teams use Django + IPython + Postgres plus Segment and Stitch to centralize data, run… | ||
| A Year of Pyxley: My First Open Source Adventure | Nicholas Kridler | 27m |
A Flask‑powered dashboard stack named Pyxley combines Flask back ends, React.js and D3 front ends and Python wrappers that take pandas DataFrames and emit… | ||
| AutoDocish: Automated-ish Dataset Documentation | Elizabeth Wickes | 30m |
A practical demo of AutoDocish claims a working, editable pipeline: a command line Python tool that reads CSVs, emits per-file Markdown profiles and a combined… | ||
| Using Twitter's Breakout Detection Package | Eric Bunch | 34m |
Twitter's breakout detection package finds abrupt sustained level shifts in univariate series, originally in R then ported to Python, and sees work on S&P 500… | ||
| Machine learning techniques for data cleaning | Cathy Deng | 24m |
Open-source Python libraries actually parse messy real-world fields: an address model uses conditional random fields to split tokens, probablepeople and… | ||
| Keynote: Developing Communities to Develop Themselves | Matthew Turk | 37m |
Describes a concrete project called YT (yours truly) used in astronomy, who writes and who uses the code, and how the community shifts from developer‑users to… | ||
| Keynote | Katy Huff | 30m |
Scaling human learning is the core problem: grassroots peer groups like the Hacker Within run local hands‑on sessions of ~30 people, while Software Carpentry… | ||
| Creating a Contemporary Risk Management System Using Python | Piero Ferrante | 47m |
Claims a working underwriting and portfolio system built on the PyData stack that uses five years of nightly invoice feeds to forecast 30/60/90 day cash flows… | ||
| Evolutionary Algorithms Perfecting the Art of "Good Enough" | Liz Sander | 30m |
Claim: evolutionary and heuristic optimizers do not guarantee the true optimum but reliably find pretty great solutions for combinatorial problems where 60 or… | ||
| Using Exploratory Data Analysis to Discover Patterns | Mehrdad Yazdani | 37m |
Develops an EDA pipeline that extracts features from images and documents using pre‑trained deep nets (ImageNet/Caffe model zoo) and text features (tf‑idf)… | ||
| When Worlds Collide: Productionalizing a Data Science Model | Tudor Radoaca, Nicole Carlson | 29m |
Built a production reliability score for Specialists with two components (retention and completion), persisted histories and a daily gauge visible in the app… | ||
| Keynote: Built in Super Heroes | David Beazley | 42m |
Argument: Python’s real power is its built-in types and small stdlib tools not fancy libraries. Uses csv module plus tuples lists sets dicts namedtuple Counter… | ||
| Finding Driving Style Patterns in Caterpillar Machine Data | Benjamin Hodel | 45m |
A working pipeline extracts engineered powertrain features from 10‑minute windows of Caterpillar telematics, reduces 50–60 features to ~8 principal components… | ||
| Implementing distributed grid search for deep learning using scikit learn and joblib | Mike Heilman | 39m |
Grid search is implemented as a distributed pipeline that wraps a TensorFlow deep text classifier in the scikit-learn estimator API and parallelizes… | ||
| Deploying Machine Learning using sklearn pipelines | Kevin Goetsch | 43m |
Sklearn pipeline objects provide a reproducible, modular workflow that encodes feature extraction, imputation, transformers, ensembles and grid search into one… | ||
| Pyglmnet: A Python package for elastic net regularized generalized linear model | Pavan Ramkumar | 45m |
An open source Python package implements elastic net regularized generalized linear models as a practical alternative to R glmnet, covering Poisson regression… | ||
| Fizz Buzz in Tensorflow | Joel Grus | 33m |
A neural network implemented with TensorFlow/Keras tries to learn the Fizz Buzz rule and succeeds unevenly: using binary encoding as input a one‑hidden‑layer… | ||
| Keynote: Using Data Science for Social Good: Examples, Opportunities, and Challenges | Rayid Ghani | 47m |
Shows concrete deployments and numbers: off‑the‑shelf machine learning on police records cuts false positives by about 50% fewer false positives while finding… | ||
| It's Not Magic: Explaining Classification Algorithms | Brian Lange | 43m |
Survey of about ten concrete classifiers with code and concrete tradeoffs: linear discriminant and logistic regression give linear boundaries and… | ||
| Adding Image and Voice Intelligence to Your Apps | David Giard | 44m |
Microsoft Cognitive Services APIs get added to apps via simple REST calls or language SDKs and return JSON with concrete outputs: face boxes and ages, emotion… | ||
| Forecasting with the Kalman Filter | Mike Mull | 39m |
Introduces the Kalman filter as an implementable forecasting engine using the development statsmodels state space code and a public GitHub notebook, fits local… | ||
| What Data Analysts Wish Application Developers Knew | Alison Stanton | 28m |
Concrete recommendations and failure cases: missing intermediate snapshots prevents save distinctions for reconciliation so demo and trial metrics diverge… | ||
| High Frequency Trading in MMORPG Markets using Luigi, Pandas, and Scikit learn | J. Henry Hinnefeld | 33m |
A Python pipeline harvests EVE Online market data via the game API for about 2,500 item types making ~5,000 calls, Luigi parallelized 20 workers to fetch and… | ||
| Data Engineering Architecture | Rob Story | 34m |
Simple runs a Redshift-centered stack with S3, Kafka (moving off RabbitMQ), PostgreSQL sources and bespoke services called Horizon and Saltooth to batch and… | ||
| Mind the Gap! Bridging the pandas – scikit learn dtype divide | Tom Augspurger | 38m |
Claim is that pandas holds labeled, heterogeneous, messy tabular data while scikit-learn expects homogeneous NumPy arrays, and that mismatch breaks workflows… | ||
| Don't Live Patch Your CPython Interpreter | James Powell | 45m |
A pip‑installable C extension called ratch live‑patches CPython by assembly‑trampolining the py_eval_frame_x loop and injecting callbacks (get_rat/set_rat) so… | ||
| Deconstructing Feather | Bill Lattner | 26m |
Claims that Feather is a cross‑language, columnar file format (Wes McKinney Hadley Wickham) that preserves types and metadata and enables zero parsing. Demo… | ||
| A Quickstart Guide to PyMC3 | Nicole Carlson | 57m |
Step‑by‑step code for building and sampling linear, hierarchical linear, logistic and hierarchical logistic models using PyMC3 with the Theano backend… | ||
| Building a Recommendation Engine with Neo4j and Python | Kevin Van Gundy | 01h 11m |
Builds a real meetup.com recommendation engine using Neo4j to glue data, Cypher for graph queries and Python for analysis and pre/post processing; imports use… | ||
| Popping Kernels An Exploration of Kernel Development for Jupyter Notebooks | Safia Abdalla | 42m |
Explains what a Jupyter kernel actually does and shows the messaging spec down to five sockets (shell, iopub, stdin, control, heartbeat), the JSON message… | ||
| Using Dask for Parallel Computing in Python | Skipper Seabold | 01h 22m |
Claims that Dask scales numpy and pandas via familiar structures and delayed task graphs and shows working examples: a billion 32‑bit floats written as 4 GB on… | ||
| Pandas: .head() to .tail() | Tom Augspurger | 01h 26m |
Hands-on pandas tutorial that actually shows what is done: cloning the GitHub repo, running Jupyter notebooks, using pandas data structures and NumPy… | ||
| Luigi & Data Pipelines | Hunter Owens | 24m |
Luigi actually shows how to turn Jupyter work into reproducible pipelines by declaring Luigi tasks and targets that map sources to outputs so runs resume from… | ||
| Pomegranate: fast and flexible probabilistic models in python | Jacob Schreiber | 01h 17m |
pomegranate claims a fast, flexible Python probabilistic toolkit that beats scikit-learn in benchmarks by using an aggressive C/Cython implementation and… | ||
| Intro to Julia | Huda Nassar | 01h 56m |
Julia claims the goal to write code like Python and let it run like C and demonstrates that with LLVM and machine code inspection, the macro code_warntype, and… | ||
| Learning scikit learn - An Introduction to Machine Learning in Python | Sebastian Raschka | 01h 27m |
Hands-on introduction to scikit-learn with runnable notebooks and a GitHub repo, showing a simple linear regression predicting brain weight from head size… | ||
| nothing matches | ||