Connecting Eastern Europe
The goal of EEML is to strengthen the Machine Learning community in Eastern Europe (EE) and to improve the diversity in the Machine Learning field in terms of gender and affiliation. We will discuss about the importance of diversity in a subsequent post. In this post we will focus on the EE community -- past, present, and future.
The (lack of an) EE identity
There are "almost as many definitions of Eastern Europe as there are scholars of the region" . That is because the countries in the region tend to have very different historical, religious, cultural, and geographical backgrounds. People speak very different languages (with latin roots, or slavic, or germanic), they have different faiths, the countries have different currencies. This rich diversity makes the region very appealing from a cultural and social perspective. But it’s the very same diversity that makes it very difficult to define an EE identity and foster a sense of community when it comes to creating a joint effort to advance research and education in the region in general and in ML/AI in particular. The fact that the educational and research systems in each country might be aligned to different standards does not help; on the contrary, it reduces to a minimum the mobility of students and research collaborations between EE countries. At the same time, the best ranked universities in Eastern Europe, according to Times Higher Education, are considerably below universities in Western Europe / US / Canada. These facts combined fuel the unprecedented brain drain phenomenon from EE countries towards the West. The existing research groups in the region tend to be quite isolated and their presence in the international publishing scene is sparse.
Current technological progress and potential impact on EE
The field of ML/AI is currently expanding very fast and advances in the field are starting to propagate very quickly into everyday life through e.g. smart phones, smart assistants, search engines, medical services. For these solutions to generate maximum positive impact, there needs to be a strong local community that has the knowledge to adapt existing solutions and develop new ones tailored for the region, and not just copy-paste solutions developed around the main AI hubs of the world. Machine learning techniques suffer from biases introduced by those developing the technology. They are learning systems that are influenced by the data they are trained on, and there is also fundamentally a bias in the type of questions that are being asked and the type of solutions that are deemed sufficient. This bias problem is not specific to only ML systems, e.g. the bias issue related to Shirley cards for skin color in photography has been around for decades, what has changed though is the speed at which we are adopting such technologies. To have a chance to correct such biases, people from all backgrounds and population groups have to be actively part of the process. Lastly, it is important to educate the population about the capabilities and limitations of this technology, to understand which things are possible and which are not (e.g. with respect to fake news, video alteration, etc.), for people to not feel threatened by the technology and preemptively reject it as a consequence.
A better future for EE with AI
EE countries are generally considered as ideal places for software outsourcing, having a strong IT literacy. For these skills to become even more relevant in the AI context, they need to be augmented with knowledge of Machine Learning, hence an imperative need for more undergrad courses and Master’s programs in ML and AI. There is a tendency to underestimate the complexity of the field, particularly due to the amount of free online materials. However, self-learning ML to the point of being truly innovative is unlikely to happen in a vacuum. It requires experience, communication, and multiple points of view. For this reason, connecting research labs and encouraging mobility is paramount. Whilst there are already some connections between research centers from EE with those from the US, Canada or Western Europe, and more such connections are welcome, the imbalance in expertise might lead to sub-par outcomes. A more sustainable way to develop expertise in the local EE community should also connect EE centers among themselves in order to build a stronger EE identity. This would encourage the exchange of students and research collaborations between these countries, contributing to creating and maintaining the talent in the region. Such an active local community would improve internationally the credibility of the countries involved in the coming years, and help the region to attract more funds to build local excellence centres in AI, that promote research and innovation in the field. Through technology, EE has the opportunity to grow economically and become an attractive hub for education and innovation.
EEML is committed to get actively involved in connecting research labs in the region and facilitate knowledge sharing through various initiatives like the EEML summer school and the EEML workshops series. The goal is to make the summer school one of the top-quality summer programs in the field, always located in one of the very welcoming Eastern European countries. It aims to bring together some of the most renowned experts in the field and students from all over the world in a fun and friendly environment, to support and highlight the local research community. Meanwhile, the workshop series consists of smaller events, 2-3 per year, spread across EE, and dedicated mainly to popularise the topic, attract students to the field, and help people understand the potential and limitations of AI.
If you have any suggestions or ideas on how to strengthen the EE community, or would like to host an EEML workshop, we would love to hear from you. Please get in touch at email@example.com or firstname.lastname@example.org!
 "The Balkans", Global Perspectives: A Remote Sensing and World Issues Site. Wheeling Jesuit University/Center for Educational Technologies, 1999–2002.If you have any comments or suggestions about this post, please write to us at email@example.com.
All roads lead to Bucharest in the summer of 2019 (08/01/2019)
After Transylvania, we take the road to Bucharest, the capital of Romania, for the 2019 edition, where we are kindly hosted by Politehnica University of Bucharest (UPB), Faculty of Automatic Control and Computer Science. Our hosts present themselves and the local AI community.
Marius: I am an Associate Professor at UPB and Senior Researcher at the Institute of Mathematics of the Romanian Academy (IMAR). At UPB, I introduced in 2014 the courses of Computer Vision and Robotics and at IMAR I started the Computer Vision Reading Group, with weekly meetings since 2016. I did my PhD at the Robotics Institute of Carnegie Mellon University and for my work on unsupervised learning for graph matching I received, in 2014, the “Grigore Moisil” Prize - the most prestigious award in Mathematics given by the Romanian Academy. Currently, I lead research projects on various computer vision and deep learning problems, e.g. unsupervised learning in the spatiotemporal domain, video to language translation, object tracking, semantic segmentation, vision for drones and self-driving cars, vision for the wood industry. Together with my students, we aim to understand vision at its deepest levels, from learning about concepts in an unsupervised fashion to understanding its relationship to natural language.
Traian: I am an Associate Professor at UPB and co-founder of the ML reading group and the Bucharest Deep Learning meetup. I hold a PhD awarded by UPB in 2012 with a thesis on discourse analysis for multi-party dialogues in an educational setting (e.g. generated in MOOCs). Since then, my research interests switched to NLP applied in question answering, code generation, people search, and conversational agents. I have worked in several European R&D projects in NLP and robotics, leading an EU accelerator grant for startups and co-leading with Marius a €1M innovation grant to develop autonomous drones powered by computer vision and language interaction.
Elena: I am a Computer Vision researcher and I lead the Theoretical Research team in ML at Bitdefender - one of the top cybersecurity companies worldwide, founded in Romania and having innovation as a core value. My research interest is in understanding videos in an unsupervised manner, currently working on object tracking for my PhD at the University of Bucharest (UB) and IMAR. I have a strong background in Mathematics and Physics and I have finished my BSc in Computer Science and the MSc in Distributed Systems at UPB.
Gabi: I am a software architect at Feel IT Services passionate about AI and the web. I am an ambassador for Iasi AI, the artificial intelligence community of Iasi where I coordinate the open data initiative and I am involved with the smart city project. I was a volunteer in the GovItHub project - a governmental initiative aimed at modernising and optimising the current government processes, both at the global and at the local level. I currently continue the projects started in GovItHub through the Civictech NGO. I have a long background of amateur AI and maker projects as well as having organised multiple international academic events.
Bucharest and its AI community
Bucharest (a.k.a. “Little Paris”), has a vibrant ML community. Its top universities, UPB and UB, offer Master’s and doctoral programs in AI. UB organises yearly a conference around advancements in AI (RAAI). Informal meetings on AI topics (meetup style) are run regularly. And there is a fast-growing startup scene, with many of these startups employing ML in their core products, e.g. FotoNation, Sparktech, Arnia, UiPath.
The interest for ML and AI was rekindled 5 years ago (in 2014), when a number of UPB faculty members and students (ourselves included) created a reading group around ML topics, with focus on Deep Learning. This reading group evolved into one of the largest research-focused communities in Romania - the Bucharest Deep Learning meetup, which now has close to 1000 members from industry, academia, and other passionates about technology. While the ML community began to evolve, more focused groups developed. We now have a Computer Vision reading group meeting regularly at IMAR, an NLP seminar at UB, an RL reading group at UPB, and a Maths for ML seminar at IMAR. In addition, the fast-evolving Bucharest.AI community and meetup, focused on the use of AI in industry and entrepreneurial life, complete the scene.
In academia, the Master’s programs at UPB and UB include courses on Robotics and Computer Vision (taught by Marius), ML (Traian), NLP, RL, and DL. Research projects around robotics, autonomous cars and UAVs, computer vision, and NLP are underway, funded mainly by the EU (structural funds and Horizon 2020) and the Romanian Government, but also by a small number of private companies. Master and PhD students publish in top conferences mainly in Computer Vision and NLP; e.g. publications list. IMAR also carries out research in fundamental ML and Computer Vision. In industry, several companies created ML teams, e.g., Bitdefender’s team of Theoretical Research in Machine Learning, BRD's research grants for PhD and postdocs working on applied ML projects in banking and finance.
Partner institutions and venue
UPB is the most prestigious technical university in Romania, with a 200 years experience in high-quality teaching. It carries research in various technical fields, including artificial intelligence. The Romanian Association for Artificial Intelligence (ARIA) was founded in 2011 by academics from several universities, in an effort to connect the different research groups in Romania among themselves and with the European scene. ARIA organises regularly events around Robotics and AI. Together, UPB and ARIA will provide the logistics support necessary in organising EEML 2019: lecture rooms and labs (at Precis Research Center), university dorms, admin support.
Irrespective of affiliation (academia, industry, meetups etc.), our purpose is to build a strong ML community here in Bucharest, mainly by: working together to consolidate our fundamental knowledge, keeping up to date with new groundbreaking results in ML through the various reading groups, and targeting top conferences to stay connected with and get feedback from the international community. We hope that, through EEML, we will contribute to building an active ML research community in Eastern Europe and overall increase our visibility globally. We are very happy that Bucharest was selected to host the 2019 edition of EEML, and we look forward to welcoming you all in Romania!
Marius, Traian, Elena, Gabi
If you have any comments or suggestions about this post, please write to us at firstname.lastname@example.org.
Empowering Eastern Europe, one step at a time (15/12/2018)
Once upon a time, a group of about 170 enthusiastic people from all over the world came together in the heart of the beautiful Transylvania, driven by their unanimous passion for machine learning. A week of intense training and motivating, even life-changing, discussions ensued. And this was only the beginning...
In July 2018, we organised the Transylvanian Machine Learning Summer School (TMLSS) as an initiative to raise the visibility of the Machine Learning community in Eastern Europe (EE) and to improve diversity and access to knowledge in the field. In a nutshell, our recipe for achieving these has three main ingredients:
(1) strengthen the communication among EE communities and between the EE and the rest of the world,
(2) correct for lack of diversity (biases) through a careful selection process, and
(3) rely on sponsorships and donations to offer low registration fees for participants and need-based scholarships.
A more detailed report on the event can be found here, and blog posts dedicated to each of these ingredients and more will follow.
Due to the success of the first edition, we are excited to announce that we will organise a new edition in the summer of 2019. Importantly, we decided to change the name of the school to better reflect our mission: Eastern European Machine Learning (EEML) summer school. The next destination will be Bucharest, the capital of Romania, and the host - Politehnica University of Bucharest, one of the top universities in Eastern Europe.
The summer school will provide a week of high-quality lectures and practical sessions given by renowned experts in the field, and will feature exciting social and networking events, some in premiere at the next edition! The participants are talented students and practitioners, selected through a careful selection process. The school welcomes applications from all over the world - not only from EE, and is open to anyone who is passionate about ML and AI - not only students. We strongly encourage candidates from underrepresented groups to apply.
The details about the application process will be published on the website in early January. The application period will open at the end of January and close at the end of March. From your application, we need to understand how the school can enhance your career and how your participation can help the school and the other participants. We want to help you succeed, and we expect you to help others succeed in return. Together, we can grow a community that values fairness and helps its members to achieve their true potential. Join us!
If you have any comments or suggestions about this post, please write to us at email@example.com.