August 2020 Newsletter

The Blosc team has just published a report on the tasks done during the present year plus a roadmap for the next things to tackle: https://blosc.org/posts/mid-2020-progress-report/

In addition, we recently released Caterva 0.4.0 (https://github.com/Blosc/Caterva/releases) and cat4py 0.4.x series (https://github.com/Blosc/cat4py/releases).  These represent a big step forward in terms of functionality and API stability.

Bokeh release 2.2 is out (blog post: https://medium.com/bokeh/bokeh-2-2-d9096a223df6)

New GSOD contributor Timo Metzger has big plans to rework Bokeh's documentation with enhanced cross-referencing and a streamlined User Guide

This year’s JuliaCon was a massive success by all accounts. With more than 10,000 registrations and over 28,900 unique viewers, this was the largest JuliaCon to date. Read more in our summary blog post: https://julialang.org/blog/2020/08/juliacon-2020-wrapup/ 

Julia v1.5.1 is now available! Check out: https://julialang.org/blog/2020/08/julia-1.5-highlights/  to learn more about the features and updates included in Julia 1.5.

The Julia Language is proud to announce that we had 5 students selected for Google summer of Code 2020. Given the number of high quality applications, we decided to supplement our 5 students supported by Google with an additional 3 students funded through our own funds. Read more here: https://discourse.julialang.org/t/the-julia-language-x-gsod-x-jsod-2020/45216 

The Julia Language’s recently created a Twitch account (https://www.twitch.tv/julialanguage) which we will be using to help grow the Julia Streamer community. Check out https://julialang.live for more info. 

Student Opportunity: If you are a student and interested in working on Open Source, check out: https://fellowship.mlh.io and make sure to indicate that you would be interested in working on Julia Projects. In the last MLH cohort, we had ~10 students working on Julia projects across the ecosystem. You can read more about one of the MLH fellows using Julia and their experience here: https://medium.com/anitab-org-open-source/how-i-benefited-from-contributing-to-open-source-ca31c168b772 

MathJax just released version 3.1.0 of MathJax.  This is a feature release that includes a number of API improvements, new extensions, updates to the assistive tools, and more examples of how to use MathJax in node applications.  It also includes several bug fixes.  See the release notes for details.

MathJax version 2.7.9 was also released, which updates the version of the Speech-Rule Engine (SRE) that underlies MathJax’s accessibility features, and adds localizations for SRE in German and French, along with access to the Clearspeak rules via the MathJax contextual menu.  See the release notes for details

MDAnalysis released its 1.0.0 release after almost 12 years in 0.x status. The 1.x releases will still support Python 2.7 because scientific projects remain that rely on a 2.7 software stack. However, 1.x will only receive major bug fixes. Primary development is taking place for the upcoming 2.0.0 release that will only support Python >= 3.6, as outlined in our Roadmap https://www.mdanalysis.org/2019/11/06/roadmap/.

We also published our Interoperability Roadmap https://www.mdanalysis.org/2020/08/03/roadmap in which we describe our efforts and plans to become API compatible with other major computational physics and chemistry Python packages.  This is a major step forward in a field that traditionally relied on file-format compatibility. This new direction was catalyzed and encouraged through discussions at the 2019 NSF MolSSI Workshop on Molecular Dynamics Software Interoperability https://molssi.org/2019/07/29/molssi-workshop-molecular-dynamics-software-interoperability/.

Released ensmallen 2.14. https://github.com/mlpack/ensmallen/releases/tag/2.14.1

All our GSoC students passed their second evaluation and are finishing up their projects successfully.  Results will be announced soon (along with another mlpack release that incorporates their great work).

The PALISADE community is hosting a monthly webinar on the use of PALISADE.  We have the webinar on the last Friday of each month.  Information about our webinar can be found here:

https://palisade-crypto.org/webinars/  

Duality Technologies is hiring senior software engineers and cryptography scientists to work on PALISADE.  Information on the positions and contact information can be found here:

https://dualitytech.com/jobs/

The PyMC dev team is excited to announce the first PyMCon, An asynchronous-first virtual conference for the Bayesian community, tentatively scheduled for the first week of November, 2020. This announcement also serves as an RFP for talks and tutorials, as well as a call for sponsorship and volunteers. You can find out all the details at the PyMCon website: https://pymc-devs.github.io/pymcon/

We hope you can participate!

rOpensci has released its first rOpenSci Community Contributing Guide! The purpose of the guide is to welcome people to rOpenSci and help them recognize themselves as potential contributors. It will help them figure out what they might gain by giving their time, expertise, and experience; match their needs with things that will help rOpenSci’s mission; and connect them with resources to help along the way. We hope it's also a useful blueprint for other open source communities. Thank you to NumFOCUS for supporting this project with a Small Development Grant.

Want to keep up with rOpenSci's work and learn what others are doing with our tools? Read and subscribe to our newsletter! A digest of R package news, use cases, blog posts, events, curated every two weeks.

Optuna 2.0, the next major evolution of Optuna released! 2.0 introduces hyperparameter importance to see which hyperparameters matter most, Hyperband pruning, and many performance improvements!

Optuna 2.0 includes the following stable features:

• Hyperparameter importance
• Grid search
• LightGBM Tuner
• Hyperband
• Built-in CMA-ES

and several experimental features, such as integration with TensorBoard and AllenNLP.

For more details, please see our blog.

The Policy Simulation Library (PSL) is a collection of open-source models and software for public-policy decisionmaking.

Recent releases:

• US federal business tax policy: Cost-of-Capital-Calculator adds additional flexibility for treatment of capital gains at death.
• US fiscal policy: OG-USA’s documentation got a makeover using Jupyter Book.
• US unemployment insurance: UI_Calculator, a new incubating project at PSL, calculates the COVID unemployment insurance benefits under the CARES Act.

General announcements:

• Analysis of Joe Biden’s tax plan (using OG-USA and Tax-Calculator) is featured by The Wall Street Journal and The Committee for a Responsible Federal Budget. Analysis of COVID relief measures’ effect on poverty and inequality (using Tax-Calculator) is presented at the Tax Economist Forum.
• The Policy Change Index (PCI)'s capabilities of predicting China’s next moves are featured in the Building the Future podcast via the Center for Strategic and International Studies.

Upcoming events:

• PCI will be demonstrated at the upcoming Data Science Conference on COVID-19 and the annual meeting of the American Political Science Association.

Next month’s PSL community calls: 3-4 p.m. ET on Sep 1, 15, and 29. Come join our calls and contribute to the various projects!

Released pysal 2.3.0 that includes a new package accessibility providing classical and novel measures of spatial accessibility to services.

Presented the tutorial Spatial Data Analysis with PySAL at SciPy 2020.

PyTorch-Ignite is a high-level library to help with training and evaluating neural networks in PyTorch flexibly and transparently.

We recently released our 0.4.1 version. The focus is on bug fixes, a few new features and the consolidation of our distributed helper module.

To learn more about the project and how to start using it, please, checkout our getting started section.

Since recently, we provide various docker images for vision/nlp tasks empowered with the latest PyTorch, PyTorch-Ignite, Nvidia/Apex and others task related libraries. More docker images with, for example, Horovod support or DeepSpeed will be available in future.

Trains Ignite server is open to everyone to browse our reproducible experiment logs, compare performances and restart any run on their own Trains server and associated infrastructure. Many thanks to the folks at Allegro AI who are making this possible!

The project is currently maintained by a team of volunteers and we are looking for motivated contributors to help us to move quickly forward. Please see the contribution guidelines for more information.

PyTorch-Ignite is presented to you with love by the PyTorch community!
Checkout the project on Github and follow us on twitter. For any questions, support or issues, please [reach out to us. For all other questions and inquiries, please send an email to contact@pytorch-ignite.ai

QuTiP 4.5.2 was released on PyPI and Conda Forge on the 14th of July, 2020, bringing support for SciPy 1.5.0, bug fixes in eigenvector calculations on macOS, the stochastic solvers and the Windows build process, and some minor speedups.

Three 2020 Google Summer of Code students are completing their projects: Jake has made an overhaul of the data structures used to store quantum objects the data layer, Sidhant has extended the integration of the quantum information processing sub-package, improving the execution, and Asad has crafted a plugin that makes Hamiltonians and other quantum objects auto-differantiable with JAX.   

We have activated a remote mentorship opportunity of 3 months for student projects, all-year-round, in collaboration with Unitary Fund. 

We have begun work on QuTiP 5.0, which will bring a whole host of improvements across the library.  Several packages are being split out into their own repositories to help manage the growing codebase, while user-facing options are getting some much-needed attention. This next major version will see a total overhaul of the data structures used to store quantum objects, letting the library seamlessly mix data representations and making it much easier for users to extend the library with new operations to suit their needs. We hope to open a public beta test in the first half of 2021, while development continues on the 4.X branch.

signac has a number of updates this month:

• Google Summer of Code students Hardik and Vishav shared their final blog posts about their work on aggregation in signac-flow and an improved design for buffering/caching data in signac.
• The signac framework was featured in "An Open Software Approach for Reproducible Research for Materials Design," an entry in the 2020 MIDAS Reproducibility Challenge. Watch the recording on YouTube.
• Alyssa Travitz will present "signac: a simple, open-source, data management framework," at the Women+ Data Science monthly webinar, a joint series between Michigan State University and the University of Michigan, on September 24 from 4-5 pm Eastern Time.

Learn more on the signac website or framework documentation, and follow @signacdata on Twitter.

Spyder released version 4.1.4 of the IDE, featuring fixes and improvements for numerous issues, along with version 0.4.0 of Spyder-Terminal with many major new features. The project also went live with its brand new docs for Spyder 4, along with a tutorial video on managing Spyder and packages with pip, Anaconda, and Conda-Forge. The team broadcast its second Spydercast episode on contributing to the IDE, and presented at SciPy 2020.

Statsmodels 0.12.0 has been released. 

Major new features include:

• New exponential smoothing model: ETS (Error, Trend, Seasonal)
• New dynamic factor model for large datasets and monthly/quarterly mixed frequency models
• Decomposition of forecast updates based on the "news"
• Sparse Cholesky Simulation Smoother
• Option to use Chandrasekhar recursions
• Two popular methods for forecasting time series, forecasting after STL decomposition and the Theta model
• Functions for constructing complex Deterministic Terms in time series models
• New statistics function: one-way ANOVA-type tests, hypothesis tests for 2-samples and meta-analysis.

239 issues were closed in this release and 221 pull requests were merged. 

Release notes with the list of changes are at https://www.statsmodels.org/stable/release/version0.12.html