#20 Paris Women in Machine Learning & Data Science: Ecological Interaction Networks, Statistical and Stochastic Models & Education
😇 We got lucky to be hosted by the CRI to celebrate our 20th meetup in style! After 2 years of existence, WiMLDS Paris counts more than 3,000 members. It was time for us to look back at what we have achieved and shared insights with our community:
WiMLDS Paris: What did we learn from our data science salary survey?
Data scientist has been called the sexiest job of the 21st century, and companies compete for top talent with generous…
🐍 We shared all the material from our coding workshop on CmdStanPy led by Mitzi Morris. If you are into Bayesian models, open source and Python, don’t miss it.
To start the evening, Raphaëlle Momal-Leisenring, PhD candidate in Applied Mathematics at AgroParisTech introduced her work on Reconstruction of Ecological Interaction Networks”.
🐠 The behavior of ecological systems mainly relies on the interactions between the species it involves. In many situations, these interactions are not observed and have to be inferred from species abundance data. Here is presented a generic statistical model for network reconstruction based on abundance data. Dependence among species is modeled by a matrix Z and in order to constrain dependencies between species a sparsity structure is forced on Z. The chosen sparsity structure is a mixture of trees. The optimization is done with the EM algorithm and is provided in this R package.
The obtained sparse structure on the dependencies among species is essential because it provides a simple explanation and allows biologists to conduct experiments (that would be too complex if the dependency structure was dense).
For the industry talk, we listened to Sarah Soleiman-Halevy, PhD candidate at Meilleurs Agents talking about “From spatial complexity to real estate prices: statistical and stochastic models”.
The challenges when working on real state data are:
- Heterogeneity: Every apartment/house is unique
- Scarcity: There is low turnover on the market
🎯 In order to build a model to estimate pricing, Sarah suggested to leverage social-spatial patterns. The main point of the talk was on clustering data while preserving topological properties. She presented Self Organizing Maps (SOM). SOM is a type of neural network applying unsupervised competitive learning. The input space is “mapped” into a two-dimensional grid (such as a rectangular grid, but other shapes are possible). The grid is composed of neurons. Similar individuals in the initial space will be projected into the same neuron or, at least, in neighboring neurons in the output space (preservation of proximity).
To conclude our meetup, we organized a panel discussion on the topic “Education: When Machine Learning Helps Human Learning”. The panel discussion was led by Marie Sacksick, PhD candidate at Domoscio.
💻 Machine Learning can rely on the data to help the teacher to spot school dropout for example, but it can also create a meta-knowledge thanks to the models: it helps to understand better human learning.
That technology will help the teachers in their everyday tasks : they will no more need to do QCM-post-processing which is not the task where they have the most added value.
One key take-away was that learning is a social activity: this is why machine learning is often embedded in a robot to be the learner’s companion.
If you want to keep posted about our activities, you are welcome to
📑check our Google spreadsheet if you want to speak 📣, host 💙, help 🌠
📍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 >email@example.com
🔥 Feel free to share your company or lab’s job positions for free on WiMLDS’ website.
A special thanks to Jean-Marc Sevin for having been an awesome host and Marie Sacksick for helping us to write this blogpost!