Doctoral Schools WUT

Search Engine for Promoters and Research Areas

Wykaz obszarów badawczych związanych z tagiem Uczenie-glebokie:

# Obszar badawczy Dziedzina naukowa
1 The research area covers the issues of image processing, computer vision, machine learning (including deep learning) and broadly understood artificial intelligence.
2 The research topic includes the development of algorithms for the processing and analysis of physiological signals, including EEG, for the purpose of the detection and prediction of epileptic seizures. The research will also cover the location of the sources of epilepsy. An important aspect of the research will be the development of algorithms that can be used in medical practice. The EEG signals database collected as part of cooperation with the Medical University of Warsaw for over 50 people is available for use. The pre-processing research is expected to develop an EMG / EEA / ECG artifact elimination method. The task involves the development of effective trait extraction methods for the detection and prediction of epileptic seizures. As an important research novelty, it is worth considering the use of deep learning, including autoencoders and convolutional neural networks.
3 Speech signal analysis to detect the emotional state of the speaker. The research task includes a comprehensive speech signal analysis to identify the speaker's emotional state. Data processing algorithms include analyzing static and dynamic signal features and selecting the best ones. Machine learning algorithms, including deep learning, will be used to determine the speaker's emotional state.
4 Analysis of physiological signals to detect fatigue. The research task requires developing and constructing a test stand that enables the acquisition of various physiological signals at different stages of user fatigue. It is planned to use the following signals: electroencephalographic (EEG), electrooculographic (EOG), and audio-visual. In the next step, methods of analyzing the acquired signals in terms of fatigue detection should be developed. Data processing algorithms include feature extraction, feature selection, and machine learning (also deep learning).

Self-supervised learning (SSL) methods have shown significant potential in improving the sample efficiency of deep learning and in providing a starting point for the learning of downstream tasks. Contrastive SSL methods have become a standard pre-training approach for a range of domains such as NLP and vision. Initial results suggest that non-contrastive SSL has been able to narrow the gap to the performance levels of contrastive methods, while obviating the need for explicit construction of negative samples. The aim of this research project is two-fold: 1) investigate the extent to which contrastive and non-contrastive methods can be used in novel SSL architectures, 2) examine whether the alignment of representations in SSL can be achieved by alternative methods, such as enforcing the ability to predict an input representation from the representations of similar inputs.


A prevalent theme present in the contemporary representation learning approaches is to pre-train large foundation models on huge datasets. Such approaches utilize static datasets constructed at a particular point in time, which contrasts with the constantly changing and expanding nature of data available on the internet. The proposed research will explore a new paradigm where the training dataset is constructed on the fly by querying the internet, enabling efficient adaptation of representation learning models to selected target tasks. The aims of this research project include 1) design methods to query relevant training data and use it to adapt the representation learning model in a continuous manner, 2) make progress towards building self-supervised methods that given a description of a task, autonomously formulate their learning curricula, query the internet for relevant training data, and use it to iteratively optimize the model.