Hello Imperial RSEs!
Shortening days, lengthening shadows; it’s the time of year when we celebrate the magical, the spooky, the weird: those p-values you can’t reproduce, that 100% validation loss glimpsed fleetingly out of the corner of your eye. Welcome to the Machine Learning Issue!*
* this is a joke, mainly
Time is running out to get your Hacktoberfest t-shirt, but you can find lots of helpful info and links from our RSE event here earlier this month.
STOP PRESS Congratulations to Diego Alonso Álvarez for claiming his t-shirt with 7 days still to go! Have you got yours? Let us know!
The perennially excellent RCS Summer School took place at the end of September, but just missed the deadline for the last newsletter. Two gentlemen from Google gave us a practical introduction to Python-flavour TensorFlow and Keras, and one hundred free US dollars each to spend in the Google Cloud shop! Very generous and frankly more than my data’s worth if I’m honest, but I didn’t say so. Highlights included a digression on bias concerning whether or not flamingos are food, and the evergreen discovery that we can now distinguish muffins from puppies. To round it off Ben Glocker from the BioMedIA group in DoC presented a masterclass on medical image analysis techniques, culminating in a worked example of age regression in brains. There were no biscuits, alas, but you can’t have everything.
Some useful links arising:
29th-30th October: Data Carpentry workshop, Cardiff.
2nd November: Open Source Hackathon 2019, London. A full day hackathon focused on the open-source Python data science ecosystem.
4th November: the DSI’s 5 year anniversary lecture, by Lord Willetts, on The Future of Research in the UK
6th November: Better Science through Better Data: Springer Nature and The Wellcome Trust bring together researchers to discuss innovative approaches to data sharing, open science, and reproducible research, together with demonstrations of exemplary projects and tools, they say.
7th November: a DoC special lecture by Prof. Ron Summers entitled How close are we to routine clinical use of AI in medical imaging?
20th November: NL-RSE is organizing the NL-RSE19 conference at the Johan Cruijff ArenA, Amsterdam.
5th/6th December: Computing Insight UK 2019 will once again take place at Manchester Central.
31st March-1st April 2020: Collaborations Workshop 2020 (CW20), Belfast. The Software Sustainability Institute’s hands-on annual workshop that brings together people with an interest in research software, from across academia and industry.
You can register as a member of The Society of Research Software Engineering priced very reasonably at only £20 per year and open to all interested parties regardless of location or affiliation.
The deadline is fast approaching (29th October) for applications to the SSI Fellowship Programme 2020. Fellows are awarded £3000 to support work in the area of software sustainability and the programme is open to applicants at all levels of experience.
If machine learning with raw TensorFlow is not your idea of fun but Keras is too high-level for what you need, then TensorLayer, a modular Python wrapper library for TensorFlow, may hit the sweet spot. It’s a highly popular library, three years’ old last month, designed to assist ML developers in assembling network layers, configuring models, and managing training data and workflow. From the docs: TensorLayer is a Deep Learning and Reinforcement Learning library extended from Google TensorFlow. It provides popular DL and RL modules that can be easily customized and assembled for tackling real-world machine learning problems. Best of all it comes with lots of examples.
Each month we highlight a piece of research software that is being used or developed at Imperial. If you have a suggestion of some software that you’d like to see us feature, please email rse-committee@imperial.ac.uk.
Questioning the meaning of your existence? Look no further.
Slides from RSE team’s presentation on creating UIs in Python at the 2019 UKRSE Conference are linked from Twitter here. If you’d be interested in a repeat of this workshop in College then let Diego know!
Blog post: A Scientific Approach to Understanding and Improving How We Develop and Use Software for Research
“Theory-Software Translation” (or “Why research software engineering can be hard!”)
Python 3.8: f-strings for debugging, more concise code via assignment expressions, and more types. Also multiprocessing shared memory may be of interest to HPC users.
TensorFlow 2.0 is out!
Gooey converts command line interfaces into desktop applications: it can be used as the frontend client for any language or program, via a minimal Python wrapper.
Hydra is a framework for building Python tools (typically CLIs), easing configuration, logging, debugging, argument parsing etc.
Streamlit is a Python app framework built specifically for Machine Learning and Data Science teams. It promotes itself as a simple and powerful app model that lets you build rich UIs incredibly quickly.
DoC have an ongoing series on Explainable AI
iPr0gram, Imperial’s lecture series for people who love programming, has restarted for 2019-20.
A Software Carpentry course is scheduled at IC for 14-15th November Please get in touch with Katerina if you can help with teaching a Python session, or providing class assistance.
Registration is open for the ARCHER online training course on shared-memory Programming with OpenMP which runs through November and December.
The list of Developer Rooms have been announced for FOSDEM 2020, including HPC, Big Data, and Data Science and Open Research Tools & Technologies
Your guest editor will be digging out her Pikachu suit for this weekend’s London ComicCon and if that doesn’t tickle your fancy there’ll surely be something for everyone at the London Jazz Festival. See you there!
The Research Computing Service offers a Data Repository for long-term archiving of research data. Uploaded data sets are automatically assigned a DOI suitable for sharing and publication.
The Research Computing Service continues to run a weekly clinic for all matters related to research computing. Bring along your HPC or programming problem or just come to talk to the RSE team about their work. See the schedule for dates and locations of upcoming clinics.
That’s all for this month, and thanks to everyone who suggested links for this edition. If you’d like anything included in the newsletter, have ideas about how it could be improved, or would even like to guest-edit a future edition then drop us a line at rse-committee@imperial.ac.uk. And if you’re reading this on the web and would like to receive the next newsletter directly to your inbox then please subscribe to our RSE Community Mailing List here.
This issue of the Research Software Community Newsletter was edited by Jazz Mack Smith