#24 Paris Women in Machine Learning & Data Science: First European joint meetup, Optical Random Features & Clustering

WiMLDS Paris
4 min readMay 5, 2020


On April 22nd, we had an original meetup… Due to the confinement, it was held online, AND it was a joint meetup with Berlin WiMLDS chapter 🌐! We were very happy to join forces with them, and to have international speakers!

The introduction was made by our co-organiser, Marie Sacksick:

1️⃣ The mission of WiMLDS is to to support and promote women and gender minorities who are practicing, studying or are interested in the fields of machine learning and data science.

2️⃣ L’Oréal-UNESCO just launched its “For Women in Science” award. If you are working in physics, mathematics, or computer sciences, you can apply until May, 31st. Five outstanding scientific researchers will be designated. Each of the five Laureates will receive an award of €100,000.

Apply !

3️⃣ WiMLDS launched an open calendar to enable our members to attend all the community’s events online. The link towards the global calendar is here and below ⤵️

4️⃣ We encourage you to follow one of our past speakers, Victoire Louis, on Twitter. Indeed, every day (#1Woman1Day), Victoire shares the story of a woman in STEM and highlights the reason why her work is major in the field. Here is an example :

Dr. Amélie Chatelain, Machine learning engineer at LightOn, kicked us off with a talk on how “Accelerating SARS-COv2 Molecular Dynamics Studies with Optical Random Features”.

Molecular Dynamics is the science of following how atoms move and interact with each other. And atoms fluctuacte every 10⁻¹⁵ seconds, leading molecules to change shape at the rate of micro or milliseconds! Numerical simulations which can help to detect all the potential form of a molecule require a high computational cost, but are essential to understand biological dynamics and potentially allow for the discovery of new drugs.

The objective is to know what are the “stable” forms of a molecule. The very slight changes of atoms will sometimes lead to a large-scale structural variations, called conformational changes.

Amelie presented us diffusion maps, a non-linear dimensionality reduction technique, and a method called no-prior-knowledge exponentially weighted moving average (NEWMA), which will help on that journey. She tested it on SARS-CoV-2, responsible for the COVID19 pandemics. Combining diffusion maps and NEWMA with optical computing leads to stables forms of the molecule, fast and with low memory footprint!

If you could not attend but want to know more about that presentation, you can have a look at this extended medium post: here.

Slides from Amélie Chatelain

Anna Vlot, Doctoral Researcher at the Berlin Institute for Medical Systems Biology, took over with a talk on “Cluster-free identification of informative features in single-cell omics data”.

In our second talk of the evening Anna Vlot, a Doctoral researcher at the Berlin Institute for Medical Sytems Biology gave a talk titled: Cluster-free identification of informative features in single-cell omics data. The field of biological sequencing has advanced to be able to look within a single for genomic coding regions. Because RNA is the expressed DNA at any given moment sequencing the RNA within a cell gives us insight into the functional response of the cell at that moment.

The development of single cell omics methods provides researchers with large sources of data to investigate the development of cell identity. Current methods to identify informative features of cell identity commonly rely on clustering groups of similar cells. In this talk Anna presented a method which enables identification of biologically informative features without reliance on previous clustering.

Anna aimed at giving us insight into genes and chromatin regions that are similar to reference cells. She outlined a few different ways she did this, including iteratively select the least similar cells, selecting cells over a 2D grid, and K-means ++. Once similar cells are selected, she tested how similar the genes were using statistical significance. She repeated similar techniques to find pre-transcriptional differences, those that are generated before RNA is fully developed. Importantly she was able to identify gene regions that were similar between the single-cell.

🙌 We want to thank our speakers for the energy! Talking in front of a screen sure is not an easy challenge, but they did it with brio.

The next meetup is scheduled on the 14th of May, three days before the International Day Againt Homophobia, Transphobia & Biphobia. It will still be online!

If you want to keep posted about our activities, you are welcome to:

📑 check our Google spreadsheet if you want to speak 📣, host 💙, help 🌠

🔗 follow our Twitter account, Meetup page, Instagram or LinkedIn page

📍join our Slack channel for more discussions about machine learning, data science, and diversity in tech!

📩 send an email to the Paris WiMLDS team to keep in touch >paris@wimlds.org

🎬 follow our WiMLDS Paris Youtube channel

📸 WiMLDS has an Instagram account, a global LinkedIn page and a Facebook page!

🔥 Feel free to share your company or lab’s job positions for free on WiMLDS’ website.



WiMLDS Paris

WiMLDS Paris is a community of women interested in Machine Learning & Data Science