#12 Paris Women in Machine Learning & Data Science: Extraction from Images, Data Visualization & Salary Negotiation
We were happy to be hosted by Meritis on the 12th of March 2019 in Paris. It was an enthusiastic full house meetup!
As usual, we kicked-off the night by sharing key information about WiMLDS and the Machine Learning community in Paris:
🔥 Feel free to share your company or lab’s job positions for free on WiMLDS’ website.
👩🏼💻 Erin LeDell, WiMLDS Founder and Chief Machine Learning Scientist at H2O.ai has been interviewed by AI4ALL Team to talk about the creation of WiMLDS and the importance of role models.
⚽️ We recommend you to check a Data Challenge focusing on Sport Analytics: a PhD or Post-Doc position will be fully financed by the PSG team! To know more about the Sport Analytics Challenge, read the details here.
🗞 The Data Analytics Post is looking to interview women data scientists, engineers or researchers! Highlighting women experts in Machine Learning and Data Science is crucial to create role models. Hence, if you are interested to be featured on the website, get in touch with us via firstname.lastname@example.org 😎
Gül first explained the evolution of human body analysis tasks through time: from image classification (i.e. is there a person on that image?) to dense pose estimation (i.e. output a 3D human shape prediction from an RGB image). On her work, she focused on the volumetric inference of 3D human shape bodies. The main challenge of this task is the lack of data which mostly relies on synthetic data or 3D shapes in constrained settings. Also, there are different choices for the representation of a 3D shape: parametric (like SMPL), point cloud or voxel representation.
After a nice introduction, Gül presented the proposed method: BodyNet. BodyNet is a neural network composed of four subnetworks, each one with a dedicated task. This solution benefits from intermediate supervision. Each of the subnetworks’ results in performance improvement as demonstrated in the experiments. The main improvement came from adding multi-view re-projection loss to the volumetric 3D loss in the final subnetwork.
And also, you can check out videos following the link on Chloé’s tweet below!
We continued the evening with Anne-Marie Tousch’s presentation on Data Visualization. Anne-Marie is a Senior Machine Learning Researcher at Criteo and deals with large-scale data every day. She motivated her talk by stating that the human visual system is a very powerful tool to process data. Take a look at Gestalt theory if you want to know more! Then, Anne-Marie shared best practices and traps to avoid ✅
Traps 👺 :
- You should not trust the summary statistics alone! It is possible to create two different datasets with a totally different underlying distribution.
- Do not use pie-charts! They are misleading 😵
Best practices 👩🎓 :
- Define your goal (explore the data? explain something specific?)
- Choose an effective visual (the simpler the better)
- Find the right focus (use color/size, remove clutter)
- Check that you actually answered your question 😂 and did not get lost on the way
For more details take a look at the slides, Anne-Marie took care to add lots of useful references to books, papers, and blogs. ⬇️
We ended the night with a Salary Negotiation Workshop led by Natalie Cernecka, EdTech expert and organizer of Paris WiMLDS. Did you know that studies [1,2,3] show that only 7% of women negotiate their first salary while 57% of men do?
Natalie Cernecka explained that “negotiation” does not come as a gift but is a skill that can (and should) be learned! Her motto is “Prepare, Practice, Persevere”. If you want to know all the details you should read Natalie’s Medium article and take a look at her resources.
- write down at least 9 reasons for why you should negotiate your salary
- a role-playing game (Star Data Scientist vs Big Boss)
- write down at least 12 high-value items for you that are easy to do for the company (i.e. additional holidays, opportunity to manage, training… be creative!)
🎬If you have not been able to attend our meetup, we have good news! It has been recorded. You can watch it below ⬇️
If you want to keep posted about our activities, you are welcome to:
📑check our Google spreadsheet if you want to speak 📣, host 💙, or help 🌠
📩send an email to the Paris WiMLDS team to keep in touch >email@example.com
📍join our Slack channel for more discussions about machine learning, data science, and diversity in tech!