DataTalks HFA #11 @ Rafael

אנו מתרגשים לקיים את המיט-אפ ה-11 במספר של סדרת המיטאפים שלנו, אותו תארח הפעם חברת רפאל, ובמיקום נוצץ - ביציע הזהב של אצטדיון סמי עופר בחיפה :)

במיט-אפ הקרוב נזכה לשמוע הרצאות מרתקות על עמידות ואמינות של מערכות לומדות - הרצאה אחת על עמידות בפני מתקפות אדברסריאליות, והרצאה שנייה על עמידות בפני אי ודאות.

המיט-אפ יתקיים ב-14.3 בין השעות 18:30-20:30 ובתפריט הרצאות מאלפות, נוף למגרש ואוכל טעים (:

מחכים לראותכם! נא הירשמו מראש
צוות דאטה-טוקס חיפה 😎

We are excited to have Rafael host our 11th meetup!
This one will be hosted by Rafael at an exciting location - the VIP section of the Sammy Ofer stadium (:

In this meetup we will discuss the exciting topic of robustness of machine learning systems. Two researchers from Rafael and Technion CS will share some of their fascinating work with us, see abstracts below.

♦ Time: March 14th, 18:30 (It's the international Pie day 🤓 🥧)
♦ Location: Sammy Ofer stadium, Haifa - The VIP Gold Hall
♦ Language: The talks will be given in Hebrew
♦ Background: Basic knowledge in data science and machine learning is advised

Please RSVP in advance.
See you there :)
The DataTalks HFA crew 😎
------------------------------------------------------------------------------------------------

Abstracts for the talks:
Dr. Itay Naeh, a lead researcher in Rafael's AI department, will share some of his research works, including robustness through cognitive dissociation mitigation

Bat-Sheva Einbinder: Training Uncertainty-Aware Classifiers with
Conformalized Deep Learning
Deep neural networks are powerful tools to detect hidden patterns in data and leverage them to make predictions, but they are not designed to understand uncertainty and estimate reliable probabilities. In particular, they tend to be overconfident. We address this problem by developing a novel training algorithm that can lead to more dependable uncertainty estimates, without sacrificing predictive power. The idea is to mitigate overconfidence by minimizing a loss function, inspired by advances in conformal inference, that quantifies model uncertainty by carefully leveraging hold-out data. Experiments with synthetic and real data demonstrate this method leads to smaller conformal prediction sets with higher conditional coverage, after exact calibration with hold-out data, compared to state-of-the-art alternatives.