About the Project
Description:
To achieve the goals of energy-efficient, sustainable, and broadly deployable AI, this research will investigate general models for vision and language that minimize inference cost and their associated optimization methods, while remaining useful across a wide range of tasks. The work will focus on quantized neural networks, robust discrete latent spaces, architecture designs, building on top of modern attention mechanisms and state space models, natively compatible with discrete computations…
and processing streaming online data incrementally.
Group:
The group is called Visual Recognition Group (VRG). It is under the Department of Cybernetics, Czech Technical University in Prague (CTU), Czech Republic. CTU ranks highly in the computer vision area. The VRG group, together with other groups at the departments of cybernetics and computer science, form a large and open international scientific environment, performing research in many topics related to deep learning and AI. There are regular scientific events such as reading groups, seminars, invited talks by international speakers (e.g. Prague Computer Science Seminar, Colloquium in Pattern Recognition). The university has computation infrastructure intended for research, including local GPU servers at the department and a GPU cluster at the university level. VRG, led by Prof. Jiri Matas, focuses on basic research and applications of computer vision and machine learning. The main research areas are object recognition and retrieval, representation learning, tracking, text recognition, and minimal problems in computer vision.
Supervisor and Project:
My name is Sasha (Oleksandr / Alexander) Shekhovtsov. I am an assistant professor at CTU teaching the “Advanced Deep Learning” course (syllabus, recordings) and a member of the European Laboratory for Learning and Intelligent Systems (ELLIS), a European network of leading AI and machine learning researchers. I am generally interested in statistical methods for machine learning (for training, handling uncertainties, analysis of stochastic and generative models, etc.), optimization tools (adaptive methods, etc), architecture design and learning principles towards more efficient AI. Recently, focusing on trained quantization methods for vision and language, leading a small team in this direction.
Your Experience and Profile:
- A relevant master's degree from a technical or mathematical school. Deepened education in subjects relevant for machine learning: statistics, optimization, data analysis, signal processing, artificial neural networks, deep learning (within the program or beyond).
- Strong mathematical background or strong experience with Deep Learning:
- Ability to understand the mathematical explanation of methods / propose new problem formulations mathematically;
- Solid programming experience: ability to understand Python / Pytorch code, develop using AI agents, develop critical parts manually;
- Motivation to explore and dig into problems, critical thinking to identify weak points in the experiments and theory;
- Fluent English.
Salary
The position will be partially funded by Toyota Motor Europe (research division), and the project is in collaboration with them. You will also receive a PhD stipend / support from the university. Your total income will be about 40-45k net CZK / month. The purchasing power of this amount in the Czech Republic can be consulted here https://data.oecd.org/chart/7j34.
What are you going to do?
Your tasks will be to:
- Perform novel research towards more efficient neural networks (understanding, proposing, implementing, conducting experiments, writing paper);
- Present research results at international conferences and journals;
- Actively collaborate within the group and with researchers worldwide;
- Assist in teaching activities such as lab assistance and student supervision;
- Pursue and complete a PhD thesis within the planned duration of four years.
Application
To apply please send your CV to O. Shekhovtsov (shekole@fel.cvut.cz). Please feel free to contact me in case of any questions.