Pedram Daee

@JessicaHullman Thank you for the examination. I am very glad that you liked the papers. Looking forward to finalizing the dissertation accordingly.

Over the holidays I enjoyed reading @ped_ram's dissertation on joint probabilistic modeling of users and data in interactive ML If you like active learning & modeling users as Bayesian agents check out papers like arxiv.org/pdf/1710.04881 bit.ly/3b94D78 @samuelkaski

@JessicaHullman @AaltoPML @samikaski have a history of works on eliciting expert knowledge to improve prediction in challenging settings. Some examples: ML2017: link.springer.com/article/10.100… Bioinformatics 2018: academic.oup.com/bioinformatics… IJCAI 2019: ijcai.org/Proceedings/20…

Today I’m looking for ML/human in the loop research papers that describe eliciting knowledge from domain experts (any form, any use in model development) Any favorites?

@marytavakol @TUeindhoven Congrats Maryam :)

Why some users "lie" to an interactive AI and even achieve better results?🤔 Watch our presentation at @UMAPconf to discover the answer and its implications! 🤯 With @ped_ram Jussi Jokinen @oulasvirta @samikaski #HCI #AI #interactiveML @CSAalto youtu.be/4noJRNVK9Ro

@elisamekler For me it was the longest journey (1999) where I met my first crush April Ryan.

Genuinely curious: What was the first game you played that had a strong emotional impact on you? Before all the JRPGs in the mid-90s, mine was with Sierra's point and click adventure EcoQuest: youtube.com/watch?v=9eoUM7…

Puzzled that most academic and industry writings (e.g., on Gamasutra) seem to suggest that games have only really emotionally impacted players since the mid- to late 2000s, bar the occasional mention of FFVII?! (probably because few people wrote about games before that?)

Users are not passive feedback sources, rather they are in control, actively steering the system towards their goal. This behaviour violates system's assumptions about the user and suggests the need for novel types of user models. Demo: fcole90.github.io/interactive_ba…

How users collaborate with an intelligent system? We designed a function optimization task where a human and an optimization algorithm collaborate to find the max of a 1-d function. With @fcole90 Jussi Jokinen @oulasvirta @samikaski. preprint: arxiv.org/pdf/2005.01291… @UMAPconf

I just gave a remote keynote on "Data analysis with humans" in PAKDD 2020 and thought to share the recording. Abstract: users.aalto.fi/~skaski/PAKDD2… slides: users.aalto.fi/~skaski/PAKDD2… recording: users.aalto.fi/~skaski/PAKDD2… @pakdd_social @FCAI_fi @AaltoPML

And the final question in the form is ...

"Due to U.S. trade controls law restrictions, your GitHub account has been restricted." The correct answer is B. #openscience #discrimination #github #USsanctions

The attached requirement is from ... A) US visa application form. B) Github appeal form regarding my account being restricted due to interacting with github while being in Iran (airtable.com/shrGBcceazKIoz…).

Going to #NeurIPS2019? Visit our poster "Machine Teaching of Active Sequential Learners" with @tomi_peltola @mert_celikok @samikaski. @FCAI_fi @AaltoPML Link: aaltopml.github.io/machine-teachi…

Our researchers have done very exciting work designing #AI that understands humans' goals better. This is a crucial step towards making truly helpful AI assistants. @samikaski @tomi_peltola @mert_celikok @ped_ram #NeurIPS fcai.fi/fcainews/2019/…

(1/2) Ever tried to get an AI system to actually do stuff? That would work much better if it understood that you do have a goal. In our NeurIPS paper with @tomi_peltola, @ped_ram and @mert_celikok we give the AI learner the capability to understand that, by learning a user model

Update: It has been 100 hours since #Iran introduced a near-total internet shutdown cutting off the country from the rest of the world 📉 Digital rights are human rights: access must be restored to all Iranians #Internet4Iran #IranProtests 📰netblocks.org/reports/intern…

Update: It has been 75 hours since #Iran introduced a near-total internet shutdown leaving Iranians isolated from the rest of the world 📉 The rights to free expression and free assembly are inviolable and must be protected #Internet4Iran #IranProtests 📰netblocks.org/reports/intern…

What if the oracle (here teacher) in active learning is allowed to lie about data labels to a machine learning algorithm? Our paper "Machine Teaching of Active Sequential Learners" is accepted at #NeurIPS2019 with @tomi_peltola, Mert Çelikok, @samikaski, #AaltoPML #FCAI @CSAalto

Our 3rd plenary meeting took place in Helsinki hosted by @UnivHelsinkiCS . We thank the consortium for two fruitful days of discussions. #coadapt #activeageing

Integrating neurophysiologic relevance feedback in intent modeling for information retrieval - JASIST, March 2019 doi.org/10.1002/asi.24… #AaltoPML