top of page

Interpretable Transformation and Analysis of Timelines through Learning via Surprisability (LvS)

Writer: Ossi MokrynOssi Mokryn

Ever had that feeling when something just seems off, even if you can’t quite explain why? 🤔



Maybe it is a sudden dip in your fitness tracker stats, an unusual spike in your energy bill, or a strange shift in market trends. Our brains are wired to notice unexpected changes - we instinctively focus on what surprises us. Osnat Mokryn noticed this and asked herself (and Teddy Lazebnik with Hagit Ben Shoshan) to imagine applying that same intuition to analyzing complex time-series data. 



In our latest work, we introduce Learning via Surprisability (LvS) - a novel approach inspired by how humans detect anomalies. Instead of treating all deviations equally, LvS prioritizes the most surprising changes, preserving critical context and making patterns easier to interpret.



🔹 Why does this matter?


 Traditional methods often drown in high-dimensional data, flagging anomalies without explaining why they matter. LvS takes a more human-like approach, focusing on what truly stands out.


🔹 What can LvS do?


 We tested it on:


 📡 Sensor data with hidden anomalies


 🌍 Global mortality trends over multiple years


 🗣️ Two centuries of U.S. State of the Union Addresses


The result? LvS efficiently highlights the most important outliers, helping us see the bigger picture instead of just noise.


If you’re working with time-series data and looking for a better way to detect and understand anomalies, LvS might be the breakthrough you need.



🔗 Read more - ArXiv paper



How do you handle unexpected shifts in your data—or even in daily life?


🔹 Example from the paper of anomaly detection - detecting interpretable outliers in a dataset that depicts the main 30 death causes world wide along a period of 30 years from 1990 -- 2019.




Analysis of causes of death over 30 years: (a) A comparison of the original vectors representing causes of death over a 30-year period. Each square corresponds to the comparison of probability distribution vectors between two years, with lighter colors indicating greater similarity. The figure clearly shows that any two consecutive years exhibit a high degree of similarity in the underlying causes of death worldwide. (b) A comparison of the Surprisal Profile vectors created by LvS of the causes of death over a 30-year period. Each square corresponds to the comparison of the SP probability distribution vectors between two years, with lighter colors indicating greater similarity. Notable anomalies are clear in the years 1994, 2004, 2008, and 2010. The values within the SP vectors enable interpretation of the most surprising elements causing the anomaly in these years. Values were normalized for easier reading.


Special thanks to Alex Abbey for his help with the code in the early stages of the work!

 
 
 

6 Comments


Tannu Rawat
Tannu Rawat
6 hours ago

Delhi area escorts can assist you if you're feeling down yet want to get upbeat. Call an escort in the Delhi area if you're having trouble. For a while, their services will let you unwind and put your problems aside. Gorgeous escorts in the Delhi region offer a variety of thrilling escort services and are available on demand. Once they have your entire concentration and you pay close attention, you can anticipate having amazing sex.

Dwarka Escorts Service || Escorts Service Saket || Escorts Service in Connaught Place || Chanakyapuri Escorts || Escorts in Greater Kailash

Like

Otis Jame
Otis Jame
11 hours ago

 Critics and players Stickman Hook have generally praised its smooth controls and satisfying swing mechanics, comparing it to classic web-based physics games.

Like

Marsh Calvin
Marsh Calvin
a day ago

It seems like you could be referring to " Sprunki Game " a popular music-based game related to the Incredibox universe.

Like

Gloria Gilmore
Gloria Gilmore
a day ago

Slope is an impressive endless running game developed as an iteration of the original slope game. Guide a ball as it traverses a series of slopes, ensuring to navigate around obstacles and depressions.

Like

Hưng Ngọc
Hưng Ngọc
5 days ago

Traditional methods are often drowned in multidimensional data, flagging outliers without explaining why they matter and that's what I love about you guys, I phoodle really hope you'll have more great articles like this to share with everyone soon.

Like

© All rights reserved to Osnat Mokryn 2020

Image credits: www.freepik.com 

bottom of page