A5025704136- Generative Adversarial Networks and Image Synthesis
- Machine Learning and Data Classification
- Advanced Vision and Imaging
- Medical Image Segmentation Techniques
- Image Processing and 3D Reconstruction
- Digital Media Forensic Detection
- Metaheuristic Optimization Algorithms Research
- Evolutionary Algorithms and Applications
- Industrial Vision Systems and Defect Detection
- Computer Graphics and Visualization Techniques
- Video Analysis and Summarization
- Advanced Multi-Objective Optimization Algorithms
- Advanced Image and Video Retrieval Techniques
- Image Retrieval and Classification Techniques
- Neural Networks and Applications
- Fuzzy Logic and Control Systems
- Face and Expression Recognition
- Natural Language Processing Techniques
- 3D Shape Modeling and Analysis
- Music and Audio Processing
- Advanced Neural Network Applications
- Handwritten Text Recognition Techniques
- Advanced Graph Neural Networks
- Advanced Image Processing Techniques
- Machine Learning and ELM
ITMO University
2016-2025
MIREA - Russian Technological University
2023
A wide range of clustering algorithms exists, most them expose many hyperparameters, on which partition quality depends. Simultaneous algorithm (model) selection and its hyperparameters optimization is considered to be a sophisticated task, known according some sources as combined hyperparameter optimization. In this paper, we focus problem selecting vector simultaneously given dataset in order achieve the best quality. We propose method for proper using reinforcement learning. Instead...
The development of sulfur-containing pharmaceutical compounds is important in the advancement medicinal chemistry. Photosensitizers (PS) that acquire new properties upon incorporation groups or individual sulfur atoms into their structure are not neglected, either. In this work, a synthesis derivatives natural chlorophyll using Lawesson's reagent was optimized. Thiocarbonyl chlorins were shown to have significant bathochromic shift absorption and fluorescence bands. feasibility...
Automated text recognition is used in autonomous driving systems, search engines, document analysis, and many other applications. There are techniques to extract information from scanned documents, but arbitrary images a much harder task. Recently suggested deep learning approaches have demonstrated highquality results, they require huge amount of data achieve them. The process collecting labelling training train network costly. In this paper, we suggest an approach for automatic dataset...
In optimisation problems, it is often more convenient and efficient to approximate a target function with another function, which would be much easier compute or analyse. Such approximations are called Surrogate Functions. A surrogate can also fairly complex object itself. For example, Random Forest Regressor (RFR) used in Hyperparameter Optimisation processes. vectors define the feature space, algorithm performance considered function. Thus, RFR trained on set of examples, performance,...
Reinforcement-based simultaneous classification model and its hyperparameters selection 253 from Auto-WEKA library have been used.The proposed method has compared with the brute force search implemented in WEKA a random time budget assignment policy.The results show significant reduction of selecting proper algorithm for processing given dataset.The often produces much better than state-of-the-art automatic hyperparameter optimization tool.
The existing methods of object image processing are analyzed. Problems using the considered within framework DLP systems considered. A new method was presented that allows complex images. metric accuracy and completeness plagiarism detection used to evaluate quality developed method. Testing carried out ranking analyze ability model search for semantics. Comparative testing with based on raster neural models. advantages disadvantages were highlighted, as well options further developmentAs a...
Subject of Research. The paper deals with research clustering algorithms for hyperparameters optimization used in machine learning. Model selection problem is comprehensively studied, and the need tradeoff between exploration exploitation identified. Thus, reduced to multi-armed bandit problem. Method. presented approach simultaneous algorithm optimization. We solution Multiarmed Bandit considered Softmax- UCB1-based variants combination different reward functions. Main Results. Experiments...
Subject of Research.The paper deals with Bayesian method for hyperparameter optimization algorithms, used in machine learning classification problems. A comprehensive survey is carried out about using a priori and posteriori knowledge task quality improvement. Method. The existing algorithm setting problems was expanded. We proposed target function modification calculated on the basis hyperparameters optimized similar metric determination similarity based generated meta-features. Main...
Currently, object detection is an important task in many areas of both industrial and research work. In practice, commonly done using a publicly available model pretrained on some dataset, for example, ImageNet. The then further fine-tuned custom dataset specific to the task. One most parameters fine-tuning training set size. According Neural Scaling Laws, larger is, higher quality. However, collecting labeling frequently costly process, so smallest size that provides desired performance...
Despite recent impressive results of generative adversarial networks on text-to-image generation, the generation complex scenes with multiple objects in complicated background remains challenging; moreover, end-to-end text-toimage still suffers from poor image quality. In this work, we propose a sequential algorithm which allows synthesizing high-quality images (more than 1024x1024 pixels). The proposed approach consists location inference, key extraction, search, layout and harmonization...