Notes on the application processGeneral Information
- Make sure all the information you provide in your application is in English; this includes Research interests section, CV, abstract (if you submit one), scholarship request. The selection committee is formed of international researchers and there is no guarantee that the person looking over your application knows any other language besides English.
Research interests section
- Please use the available space (3000 characters) to tell us more about your research interests and projects. E.g. saying only "I am interested in deep learning and reinforcement learning" is not enough. Tell us more about:
- the topics you are interested in;
- your background and knowledge in the field; what kind of projects have you worked on relevant to the theme of the school;
- the connections between your interests/projects and the theme/topics of the school
- Think of this section as a recommendation letter that you write for yourself and that will play a huge role in the decisions we make.
- Submitting an abstract is optional. Why should you submit an abstract?
- it gives you the opportunity to detail your favourite project that you mentioned in the Research interests
- it gives you the possibility to stand out from the pool of candidates, increasing your chances of being selected
- it is a good learning experience especially if you are a beginner with scientific publishing
- during the school when you will present your project abstract as a poster, it will give you the opportunity to introduce yourself to the rest of participants and lecturers and build collaborations.
- When is it understandable to not submit an abstract? For example, if you are a researcher in a different field and you are only starting to look into Deep Learning and RL to apply them to your field, then it is understandable if you do not submit an abstract. In this case, make sure you provide enough info about your background and motivations in the Research interests section
- For us to be able to judge the abstracts, consider the following:
- The abstract is not a poster. This is an example of an abstract http://www.sysml.cc/doc/137.pdf. This is an example of a poster https://guides.nyu.edu/posters. For application, you submit an abstract, and if you are selected, you will then prepare a poster, print it and present it during the school.
- The abstract should be self-contained. In principle, it should include:
- Introduction (general presentation of the problem and motivation, mention of existing/related works on the problem);
- Approach (description of the method you used to solve the problem);
- Results (description of any experiments and analysis you did);
- Conclusion (what have you learnt, what are the future works);
- References (list of citations of related works).
- The abstract must not exceed two pages, excluding references.
- If you submit as abstract a project that is in early stages and you don't have results yet, develop more the first two parts, for example the literature review. When you prepare the poster, you can then include any results you obtain.
- If you submit the (short) abstract of an already published paper, please include the venue where the full paper was published and a link where the paper can be freely accessed. You are allowed to submit the short abstract (and not a 2-page abstract) only if the paper is directly connected to the theme of the school and has already been peer-reviewed and accepted for publication in a journal/conference/workshop.
- If the paper is on arxiv only (so no peer-review), you should consider submitting a self-contained 2-page abstract to give enough details.
- If the paper has been peer-reviewed and accepted, but the topic is not directly related to the school theme, you should consider submitting a self-contained 2-page abstract to explain concepts from these fields sufficiently well for experts in deep learning and reinforcement learning to understand. In particular, define before using any acronyms from fields not related to machine learning.
- If you submit the abstract of a project that does not use Deep learning or reinforcement learning, make efforts to establish connections. E.g. would RL or neural nets be applicable there? If yes, try to identify possible approaches; if no, explain the factors that prevent it (e.g. too small training datasets)
- The selection for attending the school is independent from the selection for scholarships. We have a very limited number of scholarships, so if you request a scholarship, it is possible that you are accepted for the school, but not for the scholarship.
- Scholarships are offered based on financial considerations. When requesting a scholarship, give financial reasons and not scientific ones. "I want to learn more about RL" is not a financial reason.
- If you are selected for a scholarship, the registration (covering breakfast, lunch, coffee breaks, and social events) is free and you also receive free accommodation in a university dorm, 2 persons sharing a room. Additionally, we offer you a travel grant to cover fully or partially the travel costs.
- Note that the travel grant is given as reimbursement, so you will need to have the funds to book your travel and you will recover them (fully or partially) after the school, upon providing the corresponding receipts / invoices / tickets etc.
- We recommend using Economy class for flights/trains, unless there are special circumstances.
- Due to the high number of scholarship requests, the following limits apply for travel grants:
- For participants travelling from Romania: the actual cost or 70 EUR, whichever is lower;
- For participants travelling from Europe: the actual cost or 250 EUR, whichever is lower;
- For participants travelling from outside Europe: the actual cost or 750 EUR, whichever is lower.
Notifications of acceptance
- We will release the first round of notifications on April 13, and we will inform you if your application has been "Accepted", "Not accepted", or "On waiting list". We will also inform you at this stage about the status of your scholarship request. Candidates that are "Accepted" will have to inform us before April 20 if they wish to attend the school. Without a reply from you, we will consider it as a "No" and we will assign the place to the next candidate on the waiting list. On April 23, we will give the final notification to the candidates on the waiting list.
Frequently asked questions
- Why the name TMLSS?
- TMLSS stands for "Transylvanian Machine Learning Summer School" or, if you prefer, "To Make a Long Story Short" ;). Transylvania is a beatiful mysterious region in Romania, representative for Eastern Europe.
- Everyone is welcome to apply, regardless of location or whether you are a student or not. We strive to have participants with diverse backgrounds to encourage networking between academic and industry communities from all over the world. The location of the summer school is tied to Eastern Europe to raise visibility of the local Machine Learning community.
- Participants are expected to:
- have experience in programming (preferably Python); knowledge of any modern framework for neural networks (TensorFlow, Theano, PyTorch, etc.) will come in handy though is not strictly necessary;
- be familiar with notions of Linear Algebra, Probability Theory, general Machine Learning;
- be very interested in learning about Deep Learning and Reinforcement Learning.
- Any project related to the theme of the school (Deep Learning or Reinforcement Learning) would be relevant. The project itself does not necessarily need to rely heavily on these techniques, but should connect to them (e.g. point out where they could have been used). It does not need to be novel either. For example, you can present your experience of reproducing published work, but be original and add a personal twist. If in doubt whether your project is suitable for presentation, you can send us an email at email@example.com and ask for advice as soon as possible.
- Submissions are non-archival, so you can submit the same work to another conference/workshop.
- Yes, you can submit the abstract of an already published paper if it is related to the theme of the summer school. Please specify in the abstract the venue where the paper was published.
- Programming language: Python; Deep Learning library: TensorFlow, possibly Sonnet.
- We encourage people to apply even if they don’t have experience with TensorFlow. There will be a list of exercises posted prior to the summer school that participants can work on to get familiar with the environment used during the school.
- Yes, during the application period, from 22nd of January to 30th of March you can update or edit the form. After the deadline this is not possible anymore. We do recommend you to submit a partially completed form as early as possible and update it in due time. This will reduce the pressure as the deadline comes close and ensures you are familiar with the form, in particular how to upload files. Please make sure to keep the confirmation email received when submitting for the first time, as it contains the link to edit your form. Without this link, you will need to start the application all over again.
- Yes, you can re-upload as many times as necessary. Please ensure the file is named properly every time you upload, otherwise it might not be correctly linked to your application. You do not need to update the form after re-uploading.
- If you are interested in the topics of the summer school but the fees might prevent you from attending, you can apply for a scholarship. These are open to everyone, not only students. When filling in the Application form, please give a short justification for your request. The funding will vary on a case-by-case basis and will cover fully or partially the costs of attending the school (registration, accommodation, travel).
- Scholarships are offered on financial considerations, not on merit.
During the school, participants will be exposed to both fundamental and advanced topics related to AI. By the end of the school, participants are expected to:
- be familiar with terminology used in the field (e.g., BPTT, off/on policy, etc.);
- have a good understanding of basic and some more advanced notions in Deep Learning and Reinforcement Learning;
- understand what are the open questions in the field and what are the emergent topics;
- be exposed to some of the theoretical underpinnings and latest developments in the theoretical study of deep neural networks;
- be familiar with commonly-used neural networks (e.g., ResNet, A3C, etc.);
- understand good practices for setting up learning experiments (e.g., getting baselines, hyper-param tuning, etc.);
- be able to design, train and test a neural network for a specific task using TensorFlow;
- be able to analyse and diagnose network behaviour (e.g., overfitting).
- The school will offer the opportunity to gain visibility and network with peers and top researchers in the field through social events and poster sessions.
- Yes, you are encouraged to use your own laptop during labs. There are no specific hardware requirements. Only make sure you have Google Chrome installed, which is used to access the lab environment.
- Yes, participants will receive a certificate at the end of the school, confirming their participation at the activities of the school.
- To check whether you need a Romanian visa to attend the summer school, please go here and choose, for the "Type of visas", "Other (marked C/ZA)", and then fill in your other details. Visa exemption applies, for example, to citizens of the countries of the European Union, European Economic Area or Switzerland; holders of Schengen visas with two or multiple entries, national visas or residence permits issued by Schengen Member States; holders of permanent residence permits issued by the UK or Ireland - for more details and official information, please see here.
- If you need a visa, we will send you an invitation letter.
Your question not in the list?
Please write to us at firstname.lastname@example.org.