PyData Global 2020
2020-11-11..2020-11-15
-- online --
[PyData Global 2020, PyData Global 2022, PyData Global 2023, PyData Global 2024]
| title | author | length |
|---|---|---|
| Building A Successful Data Science Team | Justin Nguyen | 25m |
Claims three concrete drivers of success: three distinguishing practices — hiring complementary roles (data scientists, data engineers, storytellers)… | ||
| Using Dominance Analysis for accurate feature importance | Shashank Shekhar | 10m |
Demonstrates Azen and Budescu’s Dominance Analysis implemented in a Python library the uploader developed and claims it produces stable, comparable importance… | ||
| Modelling the extreme using Quantile Regression | Massimiliano Ungheretti | 27m |
Shows how conditional quantiles are estimated directly using linear econometric models and non‑linear methods implemented in tensorflow or xgboost, optimizing… | ||
| TimeSeries Forecasting with ML Algorithms and there comparisons | Sonam Pankaj | 25m |
Compares concrete algorithms on a Covid19 kaggle univariate confirmed‑cases set using RMSE and shows ARIMA giving the lowest error RMSE 4000 while moving… | ||
| Seperation Of Scales | Thomas Caswell | 35m |
Argues that design succeeds when abstractions pick a precise scale and stick to it, not when a topic is prettified; demonstrates that requests.get in two lines… | ||
| Learning from your (model's) mistakes | Simona Maggio | 11m |
Claims that a secondary binary model—Model Performance Predictor—learns to flag failure cases by training a decision tree on test features labeled… | ||
| Asynchronous fsspec file operations | Martin Durant | 08m |
fsspec gains async/await compatibility for http, s3, gcs backends so code can call file APIs from asyncio and run many transfers in parallel. The… | ||
| Why I didn't use deep learning for my image recognition | Liucija Latanauskaite | 28m |
A deep learning prototype trained with a VGG16 pre-trained base on ~300 images across 46 routes reaches ~95%+ test accuracy but fails practical checks: heavy… | ||
| Building one multitask model to rule them all | Nicole Carlson, Michael Sugimura | 28m |
A PyTorch library called Octopod streamlines training by shuffling multi-data set training batches, combining ResNet image heads with a burp text encoder and… | ||
| Using EOLearn to build a machine learning pipeline | Stuart Lynn | 33m |
Claim: a reproducible pipeline using EOLearn and Sentinel-2 multispectral data can detect floating macroplastics at sub‑pixel scale by combining the floating… | ||
| Pythons in Python: Wildlife Trade Data Analysis Using Python | Anne Devan Song | 07m |
Provides a Python package that downloads and merges the CITES CSVs (over 20 million shipment records) and FAO trade data into a single clean dataframe, fills… | ||
| NLP Spanish Challenges Alternatives | Isabel Yepes | 11m |
Most NLP resources target English so Spanish suffers from missing open corpora and fragmented datasets; Spanish counts 572 million speakers and ~8% of internet… | ||
| Is Neural Network Better Than Ash at Detecting Team Rocket | Juan De Dios Santos | 28m |
Claim: a neural net beats Ash at spotting Team Rocket and explains why. Uses TensorFlow, TensorFlow.js, Google AutoML to train an edge object detector and a… | ||
| Supercharge Scientific Computing in Python with Numba | Ankit Mahato | 22m |
Numba JIT compiles a subset of Python and NumPy into machine code and shows concrete wins: a 500×500 Mandelbrot goes from ~7 s to <1 s to <100 ms after… | ||
| Climate change analyzing remote sensing data with Python | Luis Lopez | 35m |
Claims a working workflow that lets scientists skip downloads and run reproducible analysis in Jupyter notebooks against hosted remote sensing archives, and… | ||
| Whats New In Pandas | Joris Van den Bossche | 33m |
Pandas reaches a 1.0 release in early 2020 yet continues active development with >50 regular contributors. Concrete changes: rolling/groupby accepts numba JIT… | ||
| Taking Care of Parameters So You Don't Have to with ParamTools | Hank Doupe | 29m |
A library called ParamTools processes and validates model inputs, is built on marshmallow, and exposes parameter values as numpy arrays for easy integration… | ||
| Open Source Fairness | Aileen Nielsen | 35m |
Shows concrete failures and concrete fixes: algorithms using proxy variables like healthcare spending reproduce racial gaps (Science study) and… | ||
| Inventing Curriculum using Python and spaCy | Gajendra Deshpande | 25m |
A concrete pipeline builds a dataset from job postings and written syllabi, cleans text (stopword removal, lemmatization, lowercase), computes top-10… | ||
| Using Algorithm X to reanalyse the last UK general election | Alex Glaser | 31m |
Claim: Algorithm X finds exact-cover merges of 2–4 neighboring constituencies and those merged maps feed standard PR apportionment methods. Implementation uses… | ||
| Accelerating Differential Equations In R and Python | Chris Rackauckas | 36m |
Claims real speedups: the Julia-backed diffeqpy and R wrapper diffeqr give concrete benchmarks (50–100× vs SciPy, 350× vs deSolve, a 175× Pfizer case) and… | ||
| Speed Up Your Data Processing | Chin Hwee Ong | 30m |
Claims that parallel and asynchronous Python code speeds up pipelines and shows how. Uses pandas for in-memory work and the stdlib concurrent.futures with… | ||
| Crowdsource a Distributed Organizations Data Model | Christopher Lozinski | 09m |
A crowdsourced map implements an hierarchical object-graph database modelling US Green Party politicians, organizations and events and runs on a persistent… | ||
| OpeningTheBlackBox | Ben Fowler, Chelsey Meise | 29m |
Claims that models are no longer uninterpretable, showing a pipeline on a 1.2 million row credit default dataset with 346 features, excluding protected… | ||
| Better Code for Data Science | Alexander CS Hendorf | 41m |
Argues that pragmatic engineering habits — project layout, tests, packaging, dependency pinning, code review and reproducible environments — turn notebooks… | ||
| Data Processing Pipelines for Small Big Data | Esteban Gabancho | 35m |
Claim: mid-sized datasets that fit on a laptop can be cheaper and faster than full big‑data stacks. Shows concrete techniques: generator-based ETL, async I/O /… | ||
| DevOps for Science: using continuous integration | Elle O'Brien | 31m |
Continuous integration extends to ML by versioning data, running training on demand, and surfacing detailed metric reports in pull requests; the team builds an… | ||
| Skinny Pandas Riding On A Rocket | Ian Ozsvald | 33m |
25 million rows (4 GB CSV) inflate to ~10 GB in pandas memory; converting strings to categorical: 3.4 GB → 219 MB and downcasting numerics cuts RAM and speeds… | ||
| Monitoring machine learning models in production | Arnaud Van Looveren | 31m |
Shows concrete pipelines using Alibi Detect and Seldon Core plus Knative eventing to run outlier, adversarial and drift detectors on deployed models… | ||
| Indian Sign Language Recognition (ISLAR) | Akshay Bahadur | 08m |
ISLAR claims to standardize Indian sign language by recognizing alphabets and digits and handling regional variation using image processing, hand-gesture… | ||
| Building fairer models for finance | Andrew Weeks | 30m |
Uses the Home Credit Default Risk dataset (≈300k rows) and a simple logistic regression to show fairness wins without wrecking accuracy: baseline model AUC ≈… | ||
| How to review a model | Andy Terrel | 36m |
Models are framed as a high‑risk technical‑debt asset and the claim is that disciplined review practices pay that debt down. The repo contains notebooks, a… | ||
| Ten Ways To Fizz Buzz | Joel Grus | 08m |
Ten concrete Python solutions appear and what matters is the guts not the theme: a naïve 100 print statements implementation, the canonical if/elif/else, a… | ||
| Lessons From A Nuclear Core Loading Quantum Algorithm Study | Colleen M. Farrelly | 07m |
A practical study applies quantum and quantum‑inspired ML to nuclear core loading and reports specific outcomes: problems map to an Ising to QUBO mapping and… | ||
| Ensemble X Your personal strataGEM to build Ensemble | Dipam Paul | 31m |
A concrete pipeline builds an ensemble called EnsembleX that trains three pretrained networks plus a custom CNN on OCT skin lesion and malaria sets uses data… | ||
| What cyber security can teach us about COVID-19 testing | Hagit Grushka | 23m |
Claim: allocate a small share of tests to systematic exploration and the rest to high-risk ranking, and governments learn spread fast while keeping… | ||
| Enquiry Based Learning for Science and Engineering utilizing Bokeh | Brian Lucas | 10m |
Website hosts three models (reaction kinetics for sequential reactions, a zero‑energy evaporative cooling chamber, and an SCIR 8‑class infectious model). Each… | ||
| Snap ML: Accelerated, Accurate,Efficient,Machine Learning | Haris Pozidis | 30m |
Library codebase is C++/CUDA with Python wrappers following the scikit-learn API, offering CPU, multi-threaded, GPU, multi-GPU and distributed solvers and an… | ||
| Data Visualization & Storytelling | jose berengueres | 11m |
Claims that poor visuals come from education and tool misuse and backs it with concrete examples: bad charts from the web, a gender‑breakdown example that… | ||
| A crash-update to lifelines | Cameron Davidson-Pilon | 05m |
Lifelines now uses automatic differentiation to enable new statistical models and faster computation, it maintains a rapid release cadence of every two to… | ||
| Dirty Data Science Machine Learning On Non Curated Data | Gael Varoquaux | 29m |
Claims that string‑aware encodings and simple pipelines let supervised models learn directly from messy tables without manual curation, demonstrating a dataset… | ||
| Visual data abundant relevant labelled cheap Pick two | Irina Vidal Migallón | 30m |
Describes concrete tools and limits: use transfer learning, classic augmentations and use adversarial augmentations in both training and testing to probe… | ||
| An Introduction to DataFrames for pandas Users | Bogumił Kamiński | 30m |
Shows how to reproduce pandas-style DataFrame workflows in Julia using DataFrames.jl, with line-by-line notebook conversions that cover reading CSV/Parquet… | ||
| nothing matches | ||