Search

Your search keyword '"Koziarski A"' showing total 200 results
Start Over Author "Koziarski A" Search Limiters Available in Library Collection
200 results on '"Koziarski A"'

1. Cell Morphology-Guided Small Molecule Generation with GFlowNets

Author

Lu, Stephen Zhewen, Lu, Ziqing, Hajiramezanali, Ehsan, Biancalani, Tommaso, Bengio, Yoshua, Scalia, Gabriele, and Koziarski, Michał

Subjects
Computer Science - Machine Learning, Quantitative Biology - Biomolecules
Abstract

High-content phenotypic screening, including high-content imaging (HCI), has gained popularity in the last few years for its ability to characterize novel therapeutics without prior knowledge of the protein target. When combined with deep learning techniques to predict and represent molecular-phenotype interactions, these advancements hold the potential to significantly accelerate and enhance drug discovery applications. This work focuses on the novel task of HCI-guided molecular design. Generative models for molecule design could be guided by HCI data, for example with a supervised model that links molecules to phenotypes of interest as a reward function. However, limited labeled data, combined with the high-dimensional readouts, can make training these methods challenging and impractical. We consider an alternative approach in which we leverage an unsupervised multimodal joint embedding to define a latent similarity as a reward for GFlowNets. The proposed model learns to generate new molecules that could produce phenotypic effects similar to those of the given image target, without relying on pre-annotated phenotypic labels. We demonstrate that the proposed method generates molecules with high morphological and structural similarity to the target, increasing the likelihood of similar biological activity, as confirmed by an independent oracle model.

Published
2024

2. RetroGFN: Diverse and Feasible Retrosynthesis using GFlowNets

Author

Gaiński, Piotr, Koziarski, Michał, Maziarz, Krzysztof, Segler, Marwin, Tabor, Jacek, and Śmieja, Marek

Subjects
Computer Science - Machine Learning
Abstract

Single-step retrosynthesis aims to predict a set of reactions that lead to the creation of a target molecule, which is a crucial task in molecular discovery. Although a target molecule can often be synthesized with multiple different reactions, it is not clear how to verify the feasibility of a reaction, because the available datasets cover only a tiny fraction of the possible solutions. Consequently, the existing models are not encouraged to explore the space of possible reactions sufficiently. In this paper, we propose a novel single-step retrosynthesis model, RetroGFN, that can explore outside the limited dataset and return a diverse set of feasible reactions by leveraging a feasibility proxy model during the training. We show that RetroGFN achieves competitive results on standard top-k accuracy while outperforming existing methods on round-trip accuracy. Moreover, we provide empirical arguments in favor of using round-trip accuracy which expands the notion of feasibility with respect to the standard top-k accuracy metric.

Published
2024

3. RGFN: Synthesizable Molecular Generation Using GFlowNets

Author

Koziarski, Michał, Rekesh, Andrei, Shevchuk, Dmytro, van der Sloot, Almer, Gaiński, Piotr, Bengio, Yoshua, Liu, Cheng-Hao, Tyers, Mike, and Batey, Robert A.

Subjects
Physics - Chemical Physics, Computer Science - Machine Learning, Quantitative Biology - Biomolecules
Abstract

Generative models hold great promise for small molecule discovery, significantly increasing the size of search space compared to traditional in silico screening libraries. However, most existing machine learning methods for small molecule generation suffer from poor synthesizability of candidate compounds, making experimental validation difficult. In this paper we propose Reaction-GFlowNet (RGFN), an extension of the GFlowNet framework that operates directly in the space of chemical reactions, thereby allowing out-of-the-box synthesizability while maintaining comparable quality of generated candidates. We demonstrate that with the proposed set of reactions and building blocks, it is possible to obtain a search space of molecules orders of magnitude larger than existing screening libraries coupled with low cost of synthesis. We also show that the approach scales to very large fragment libraries, further increasing the number of potential molecules. We demonstrate the effectiveness of the proposed approach across a range of oracle models, including pretrained proxy models and GPU-accelerated docking.

Published
2024

4. Towards DNA-Encoded Library Generation with GFlowNets

Author

Koziarski, Michał, Abukalam, Mohammed, Shah, Vedant, Vaillancourt, Louis, Schuetz, Doris Alexandra, Jain, Moksh, van der Sloot, Almer, Bourgey, Mathieu, Marinier, Anne, and Bengio, Yoshua

Subjects
Computer Science - Machine Learning, Quantitative Biology - Quantitative Methods
Abstract

DNA-encoded libraries (DELs) are a powerful approach for rapidly screening large numbers of diverse compounds. One of the key challenges in using DELs is library design, which involves choosing the building blocks that will be combinatorially combined to produce the final library. In this paper we consider the task of protein-protein interaction (PPI) biased DEL design. To this end, we evaluate several machine learning algorithms on the PPI modulation task and use them as a reward for the proposed GFlowNet-based generative approach. We additionally investigate the possibility of using structural information about building blocks to design a hierarchical action space for the GFlowNet. The observed results indicate that GFlowNets are a promising approach for generating diverse combinatorial library candidates.

Published
2024

5. Towards equilibrium molecular conformation generation with GFlowNets

Author

Volokhova, Alexandra, Koziarski, Michał, Hernández-García, Alex, Liu, Cheng-Hao, Miret, Santiago, Lemos, Pablo, Thiede, Luca, Yan, Zichao, Aspuru-Guzik, Alán, and Bengio, Yoshua

Subjects
Computer Science - Machine Learning, Computer Science - Artificial Intelligence
Abstract

Sampling diverse, thermodynamically feasible molecular conformations plays a crucial role in predicting properties of a molecule. In this paper we propose to use GFlowNet for sampling conformations of small molecules from the Boltzmann distribution, as determined by the molecule's energy. The proposed approach can be used in combination with energy estimation methods of different fidelity and discovers a diverse set of low-energy conformations for highly flexible drug-like molecules. We demonstrate that GFlowNet can reproduce molecular potential energy surfaces by sampling proportionally to the Boltzmann distribution.

Published
2023

6. Crystal-GFN: sampling crystals with desirable properties and constraints

Author

AI4Science, Mila, Hernandez-Garcia, Alex, Duval, Alexandre, Volokhova, Alexandra, Bengio, Yoshua, Sharma, Divya, Carrier, Pierre Luc, Benabed, Yasmine, Koziarski, Michał, and Schmidt, Victor

Subjects
Computer Science - Machine Learning
Abstract

Accelerating material discovery holds the potential to greatly help mitigate the climate crisis. Discovering new solid-state materials such as electrocatalysts, super-ionic conductors or photovoltaic materials can have a crucial impact, for instance, in improving the efficiency of renewable energy production and storage. In this paper, we introduce Crystal-GFN, a generative model of crystal structures that sequentially samples structural properties of crystalline materials, namely the space group, composition and lattice parameters. This domain-inspired approach enables the flexible incorporation of physical and structural hard constraints, as well as the use of any available predictive model of a desired physicochemical property as an objective function. To design stable materials, one must target the candidates with the lowest formation energy. Here, we use as objective the formation energy per atom of a crystal structure predicted by a new proxy machine learning model trained on MatBench. The results demonstrate that Crystal-GFN is able to sample highly diverse crystals with low (median -3.1 eV/atom) predicted formation energy., Comment: Main paper (10 pages) + references + appendix

Published
2023

7. Towards Foundational Models for Molecular Learning on Large-Scale Multi-Task Datasets

Author

Beaini, Dominique, Huang, Shenyang, Cunha, Joao Alex, Li, Zhiyi, Moisescu-Pareja, Gabriela, Dymov, Oleksandr, Maddrell-Mander, Samuel, McLean, Callum, Wenkel, Frederik, Müller, Luis, Mohamud, Jama Hussein, Parviz, Ali, Craig, Michael, Koziarski, Michał, Lu, Jiarui, Zhu, Zhaocheng, Gabellini, Cristian, Klaser, Kerstin, Dean, Josef, Wognum, Cas, Sypetkowski, Maciej, Rabusseau, Guillaume, Rabbany, Reihaneh, Tang, Jian, Morris, Christopher, Koutis, Ioannis, Ravanelli, Mirco, Wolf, Guy, Tossou, Prudencio, Mary, Hadrien, Bois, Therence, Fitzgibbon, Andrew, Banaszewski, Błażej, Martin, Chad, and Masters, Dominic

Subjects
Computer Science - Machine Learning
Abstract

Recently, pre-trained foundation models have enabled significant advancements in multiple fields. In molecular machine learning, however, where datasets are often hand-curated, and hence typically small, the lack of datasets with labeled features, and codebases to manage those datasets, has hindered the development of foundation models. In this work, we present seven novel datasets categorized by size into three distinct categories: ToyMix, LargeMix and UltraLarge. These datasets push the boundaries in both the scale and the diversity of supervised labels for molecular learning. They cover nearly 100 million molecules and over 3000 sparsely defined tasks, totaling more than 13 billion individual labels of both quantum and biological nature. In comparison, our datasets contain 300 times more data points than the widely used OGB-LSC PCQM4Mv2 dataset, and 13 times more than the quantum-only QM1B dataset. In addition, to support the development of foundational models based on our proposed datasets, we present the Graphium graph machine learning library which simplifies the process of building and training molecular machine learning models for multi-task and multi-level molecular datasets. Finally, we present a range of baseline results as a starting point of multi-task and multi-level training on these datasets. Empirically, we observe that performance on low-resource biological datasets show improvement by also training on large amounts of quantum data. This indicates that there may be potential in multi-task and multi-level training of a foundation model and fine-tuning it to resource-constrained downstream tasks.

Published
2023

8. ChiENN: Embracing Molecular Chirality with Graph Neural Networks

Author

Gaiński, Piotr, Koziarski, Michał, Tabor, Jacek, and Śmieja, Marek

Subjects
Computer Science - Machine Learning, Quantitative Biology - Quantitative Methods
Abstract

Graph Neural Networks (GNNs) play a fundamental role in many deep learning problems, in particular in cheminformatics. However, typical GNNs cannot capture the concept of chirality, which means they do not distinguish between the 3D graph of a chemical compound and its mirror image (enantiomer). The ability to distinguish between enantiomers is important especially in drug discovery because enantiomers can have very distinct biochemical properties. In this paper, we propose a theoretically justified message-passing scheme, which makes GNNs sensitive to the order of node neighbors. We apply that general concept in the context of molecular chirality to construct Chiral Edge Neural Network (ChiENN) layer which can be appended to any GNN model to enable chirality-awareness. Our experiments show that adding ChiENN layers to a GNN outperforms current state-of-the-art methods in chiral-sensitive molecular property prediction tasks.

Published
2023

10. DiagSet: a dataset for prostate cancer histopathological image classification

Author

Michał Koziarski, Bogusław Cyganek, Przemysław Niedziela, Bogusław Olborski, Zbigniew Antosz, Marcin Żydak, Bogdan Kwolek, Paweł Wąsowicz, Andrzej Bukała, Jakub Swadźba, and Piotr Sitkowski

Subjects
Medicine, Science
Abstract

Abstract Cancer diseases constitute one of the most significant societal challenges. In this paper, we introduce a novel histopathological dataset for prostate cancer detection. The proposed dataset, consisting of over 2.6 million tissue patches extracted from 430 fully annotated scans, 4675 scans with assigned binary diagnoses, and 46 scans with diagnoses independently provided by a group of histopathologists can be found at https://github.com/michalkoziarski/DiagSet . Furthermore, we propose a machine learning framework for detection of cancerous tissue regions and prediction of scan-level diagnosis, utilizing thresholding to abstain from the decision in uncertain cases. The proposed approach, composed of ensembles of deep neural networks operating on the histopathological scans at different scales, achieves 94.6% accuracy in patch-level recognition and is compared in a scan-level diagnosis with 9 human histopathologists showing high statistical agreement.

Published
2024
Full Text
View/download PDF

11. Are Parents Logged In? The Importance of Parent Involvement in K-12 Online Learning

Author

Ricker, Gina, Belenky, Daniel, and Koziarski, Mathew

Abstract

Among the many factors that impact student performance in traditional K-12 settings, parent involvement (PI) is consistently identified as critical for student success. However, less is known about parent involvement for full-time virtual schools, where most learning takes place asynchronously in a Learning Management System (LMS). The present study attempts to close this gap by using data from three virtual schools' LMS to measure the impact of parent involvement on mathematics achievement, across grade bands where PI is known to vary. After controlling for factors known to impact education outcomes, parental involvement - as measured by LMS logins - had a small, but positive, impact on student performance in mathematics across elementary, middle, and high school. These results, coupled with the elevated role of the parent, suggest that parental involvement is just as critical, if not more so, in virtual schools as in traditional settings.

Published
2021

12. Imbalanced data preprocessing techniques utilizing local data characteristics

Author

Koziarski, Michał

Subjects
Computer Science - Machine Learning
Abstract

Data imbalance, that is the disproportion between the number of training observations coming from different classes, remains one of the most significant challenges affecting contemporary machine learning. The negative impact of data imbalance on traditional classification algorithms can be reduced by the data preprocessing techniques, methods that manipulate the training data to artificially reduce the degree of imbalance. However, the existing data preprocessing techniques, in particular SMOTE and its derivatives, which constitute the most prevalent paradigm of imbalanced data preprocessing, tend to be susceptible to various data difficulty factors. This is in part due to the fact that the original SMOTE algorithm does not utilize the information about majority class observations. The focus of this thesis is development of novel data resampling strategies natively utilizing the information about the distribution of both minority and majority class. The thesis summarizes the content of 12 research papers focused on the proposed binary data resampling strategies, their translation to the multi-class setting, and the practical application to the problem of histopathological data classification., Comment: PhD Thesis

Published
2021

13. Student Clickstream Data: Does Time of Day Matter?

Author

Ricker, Gina M., Koziarski, Mathew, and Walters, Alyssa M.

Abstract

The relationship between student activity data and performance in the online classroom is well-documented, yet the parameters of this relationship and their implications for K-12 online schools are not yet well understood. This study examined the role of student chronotype (defined here as the time of day a student is most active in an online course) and overall activity level on course performance. Clickstream data captured by a Learning Management System from 414 students enrolled in an eleventh-grade English course in the fall of 2018 at two Midwestern full-time K-12 virtual schools were used to determine chronotype and activity level. Students were classified as one of four possible chronotypes given the mode of their click activity. Results of an ANCOVA showed that students who were most active in the morning significantly outperformed students who were most active in the afternoon and evening. Morning students also tended to be the most active overall. The results of a hierarchical regression revealed that, while chronotype was related to student performance, total student activity had an even greater impact on performance suggesting an interesting interplay between the two factors.

Published
2020

14. DiagSet: a dataset for prostate cancer histopathological image classification

Author

Koziarski, Michał, Cyganek, Bogusław, Niedziela, Przemysław, Olborski, Bogusław, Antosz, Zbigniew, Żydak, Marcin, Kwolek, Bogdan, Wąsowicz, Paweł, Bukała, Andrzej, Swadźba, Jakub, and Sitkowski, Piotr

Subjects
Electrical Engineering and Systems Science - Image and Video Processing, Computer Science - Computer Vision and Pattern Recognition, Computer Science - Machine Learning
Abstract

Cancer diseases constitute one of the most significant societal challenges. In this paper, we introduce a novel histopathological dataset for prostate cancer detection. The proposed dataset, consisting of over 2.6 million tissue patches extracted from 430 fully annotated scans, 4675 scans with assigned binary diagnoses, and 46 scans with diagnoses independently provided by a group of histopathologists can be found at https://github.com/michalkoziarski/DiagSet. Furthermore, we propose a machine learning framework for detection of cancerous tissue regions and prediction of scan-level diagnosis, utilizing thresholding to abstain from the decision in uncertain cases. The proposed approach, composed of ensembles of deep neural networks operating on the histopathological scans at different scales, achieves 94.6% accuracy in patch-level recognition and is compared in a scan-level diagnosis with 9 human histopathologists showing high statistical agreement.

Published
2021

15. RB-CCR: Radial-Based Combined Cleaning and Resampling algorithm for imbalanced data classification

Author

Koziarski, Michał, Bellinger, Colin, and Woźniak, Michał

Subjects
Computer Science - Machine Learning
Abstract

Real-world classification domains, such as medicine, health and safety, and finance, often exhibit imbalanced class priors and have asynchronous misclassification costs. In such cases, the classification model must achieve a high recall without significantly impacting precision. Resampling the training data is the standard approach to improving classification performance on imbalanced binary data. However, the state-of-the-art methods ignore the local joint distribution of the data or correct it as a post-processing step. This can causes sub-optimal shifts in the training distribution, particularly when the target data distribution is complex. In this paper, we propose Radial-Based Combined Cleaning and Resampling (RB-CCR). RB-CCR utilizes the concept of class potential to refine the energy-based resampling approach of CCR. In particular, RB-CCR exploits the class potential to accurately locate sub-regions of the data-space for synthetic oversampling. The category sub-region for oversampling can be specified as an input parameter to meet domain-specific needs or be automatically selected via cross-validation. Our $5\times2$ cross-validated results on 57 benchmark binary datasets with 9 classifiers show that RB-CCR achieves a better precision-recall trade-off than CCR and generally out-performs the state-of-the-art resampling methods in terms of AUC and G-mean.

Published
2021

16. Potential Anchoring for imbalanced data classification

Author

Koziarski, Michał

Subjects
Computer Science - Machine Learning, Statistics - Machine Learning
Abstract

Data imbalance remains one of the factors negatively affecting the performance of contemporary machine learning algorithms. One of the most common approaches to reducing the negative impact of data imbalance is preprocessing the original dataset with data-level strategies. In this paper we propose a unified framework for imbalanced data over- and undersampling. The proposed approach utilizes radial basis functions to preserve the original shape of the underlying class distributions during the resampling process. This is done by optimizing the positions of generated synthetic observations with respect to the potential resemblance loss. The final Potential Anchoring algorithm combines over- and undersampling within the proposed framework. The results of the experiments conducted on 60 imbalanced datasets show outperformance of Potential Anchoring over state-of-the-art resampling algorithms, including previously proposed methods that utilize radial basis functions to model class potential. Furthermore, the results of the analysis based on the proposed data complexity index show that Potential Anchoring is particularly well suited for handling naturally complex (i.e. not affected by the presence of noise) datasets.

Published
2021

17. CSMOUTE: Combined Synthetic Oversampling and Undersampling Technique for Imbalanced Data Classification

Author

Koziarski, Michał

Subjects
Computer Science - Machine Learning, Statistics - Machine Learning
Abstract

In this paper we propose a novel data-level algorithm for handling data imbalance in the classification task, Synthetic Majority Undersampling Technique (SMUTE). SMUTE leverages the concept of interpolation of nearby instances, previously introduced in the oversampling setting in SMOTE. Furthermore, we combine both in the Combined Synthetic Oversampling and Undersampling Technique (CSMOUTE), which integrates SMOTE oversampling with SMUTE undersampling. The results of the conducted experimental study demonstrate the usefulness of both the SMUTE and the CSMOUTE algorithms, especially when combined with more complex classifiers, namely MLP and SVM, and when applied on datasets consisting of a large number of outliers. This leads us to a conclusion that the proposed approach shows promise for further extensions accommodating local data characteristics, a direction discussed in more detail in the paper.

Published
2020

18. Combined Cleaning and Resampling Algorithm for Multi-Class Imbalanced Data with Label Noise

Author

Koziarski, Michał, Woźniak, Michał, and Krawczyk, Bartosz

Subjects
Computer Science - Machine Learning, Statistics - Machine Learning
Abstract

The imbalanced data classification is one of the most crucial tasks facing modern data analysis. Especially when combined with other difficulty factors, such as the presence of noise, overlapping class distributions, and small disjuncts, data imbalance can significantly impact the classification performance. Furthermore, some of the data difficulty factors are known to affect the performance of the existing oversampling strategies, in particular SMOTE and its derivatives. This effect is especially pronounced in the multi-class setting, in which the mutual imbalance relationships between the classes complicate even further. Despite that, most of the contemporary research in the area of data imbalance focuses on the binary classification problems, while their more difficult multi-class counterparts are relatively unexplored. In this paper, we propose a novel oversampling technique, a Multi-Class Combined Cleaning and Resampling (MC-CCR) algorithm. The proposed method utilizes an energy-based approach to modeling the regions suitable for oversampling, less affected by small disjuncts and outliers than SMOTE. It combines it with a simultaneous cleaning operation, the aim of which is to reduce the effect of overlapping class distributions on the performance of the learning algorithms. Finally, by incorporating a dedicated strategy of handling the multi-class problems, MC-CCR is less affected by the loss of information about the inter-class relationships than the traditional multi-class decomposition strategies. Based on the results of experimental research carried out for many multi-class imbalanced benchmark datasets, the high robust of the proposed approach to noise was shown, as well as its high quality compared to the state-of-art methods.

Published
2020

19. Two-Stage Resampling for Convolutional Neural Network Training in the Imbalanced Colorectal Cancer Image Classification

Author

Koziarski, Michał

Subjects
Computer Science - Machine Learning, Computer Science - Computer Vision and Pattern Recognition, Statistics - Machine Learning
Abstract

Data imbalance remains one of the open challenges in the contemporary machine learning. It is especially prevalent in case of medical data, such as histopathological images. Traditional data-level approaches for dealing with data imbalance are ill-suited for image data: oversampling methods such as SMOTE and its derivatives lead to creation of unrealistic synthetic observations, whereas undersampling reduces the amount of available data, critical for successful training of convolutional neural networks. To alleviate the problems associated with over- and undersampling we propose a novel two-stage resampling methodology, in which we initially use the oversampling techniques in the image space to leverage a large amount of data for training of a convolutional neural network, and afterwards apply undersampling in the feature space to fine-tune the last layers of the network. Experiments conducted on a colorectal cancer image dataset indicate the usefulness of the proposed approach.

Published
2020

20. Radial-Based Undersampling for Imbalanced Data Classification

Author

Koziarski, Michał

Subjects
Computer Science - Machine Learning, Statistics - Machine Learning
Abstract

Data imbalance remains one of the most widespread problems affecting contemporary machine learning. The negative effect data imbalance can have on the traditional learning algorithms is most severe in combination with other dataset difficulty factors, such as small disjuncts, presence of outliers and insufficient number of training observations. Aforementioned difficulty factors can also limit the applicability of some of the methods of dealing with data imbalance, in particular the neighborhood-based oversampling algorithms based on SMOTE. Radial-Based Oversampling (RBO) was previously proposed to mitigate some of the limitations of the neighborhood-based methods. In this paper we examine the possibility of utilizing the concept of mutual class potential, used to guide the oversampling process in RBO, in the undersampling procedure. Conducted computational complexity analysis indicates a significantly reduced time complexity of the proposed Radial-Based Undersampling algorithm, and the results of the performed experimental study indicate its usefulness, especially on difficult datasets.

Published
2019

21. Analysis of group evolution prediction in complex networks

Author

Saganowski, Stanisław, Bródka, Piotr, Koziarski, Michał, and Kazienko, Przemysław

Subjects
Computer Science - Social and Information Networks, Physics - Physics and Society
Abstract

In the world, in which acceptance and the identification with social communities are highly desired, the ability to predict evolution of groups over time appears to be a vital but very complex research problem. Therefore, we propose a new, adaptable, generic and mutli-stage method for Group Evolution Prediction (GEP) in complex networks, that facilitates reasoning about the future states of the recently discovered groups. The precise GEP modularity enabled us to carry out extensive and versatile empirical studies on many real-world complex / social networks to analyze the impact of numerous setups and parameters like time window type and size, group detection method, evolution chain length, prediction models, etc. Additionally, many new predictive features reflecting the group state at a given time have been identified and tested. Some other research problems like enriching learning evolution chains with external data have been analyzed as well.

Published
2017
Full Text
View/download PDF

22. Multicriteria Classifier Ensemble Learning for Imbalanced Data

Author

Weronika Wegier, Michal Koziarski, and Micha Wozniak

Subjects
Classifier ensemble, imbalanced data, multi-objective optimization, pattern classification, Electrical engineering. Electronics. Nuclear engineering, TK1-9971
Abstract

One of the vital problems with the imbalanced data classifier training is the definition of an optimization criterion. Typically, since the exact cost of misclassification of the individual classes is unknown, combined metrics and loss functions that roughly balance the cost for each class are used. However, this approach can lead to a loss of information, since different trade-offs between class misclassification rates can produce similar combined metric values. To address this issue, this paper discusses a multi-criteria ensemble training method for the imbalanced data. The proposed method jointly optimizes precision and recall, and provides the end-user with a set of Pareto optimal solutions, from which the final one can be chosen according to the user’s preference. The proposed approach was evaluated on a number of benchmark datasets and compared with the single-criterion approach (where the selected criterion was one of the chosen metrics). The results of the experiments confirmed the usefulness of the obtained method, which on the one hand guarantees good quality, i.e., not worse than the one obtained with the use of single-criterion optimization, and on the other hand, offers the user the opportunity to choose the solution that best meets their expectations regarding the trade-off between errors on the minority and the majority class.

Published
2022
Full Text
View/download PDF

23. The effect of the COVID-19 pandemic on mental health calls for police service

Author

Jacek Koziarski

Subjects
Police, Mental Health, COVID-19, Bayesian Structural Time Series, Science (General), Q1-390, Social pathology. Social and public welfare. Criminology, HV1-9960
Abstract

Abstract Drawing upon seven years of police calls for service data (2014–2020), this study examined the effect of the COVID-19 pandemic on calls involving persons with perceived mental illness (PwPMI) using a Bayesian Structural Time Series. The findings revealed that PwPMI calls did not increase immediately after the beginning of the pandemic in March 2020. Instead, a sustained increase in PwPMI calls was identified in August 2020 that later became statistically significant in October 2020. Ultimately, the analysis revealed a 22% increase in PwPMI calls during the COVID-19 pandemic than would have been expected had the pandemic not taken place. The delayed effect of the pandemic on such calls points to a need for policymakers to prioritize widely accessible mental health care that can be deployed early during public health emergencies thus potentially mitigating or eliminating the need for increased police intervention, as was the case here.

Published
2021
Full Text
View/download PDF

24. Caught in the currents: evaluating the evidence for common downstream police response interventions in calls involving persons with mental illness

Author

Laura Huey, Judith Andersen, Craig Bennell, Mary Ann Campbell, Jacek Koziarski, and Adam D. Vaughan

Subjects
policing, mental health, covid-19, police response, Education, Science
Abstract

The origins of this report, and of the Mental Health and Policing Working Group, can be traced to the unique situation Canadians have faced as a result of the COVID-19 pandemic. The unique circumstances of this global outbreak, which have for many Canadians resulted in serious illness and death, intensified economic uncertainties, altered family and lifestyle dynamics, and generated or exacerbated feelings of loneliness and social dislocation, rightly led the Royal Society of Canada’s COVID-19 Taskforce to consider the strains and other negative impacts on individual, group, and community mental health. With the central role that police too often play in the lives of individuals in mental and (or) emotional crisis, we were tasked with exploring what can be reasonably said about the state of our current knowledge of police responses to persons with mental illness.

Published
2021
Full Text
View/download PDF

26. A Study on Pattern Recognition with the Histograms of Oriented Gradients in Distorted and Noisy Images

Author

Andrzej Bukała, Michał Koziarski, Bogusław Cyganek, Osman Koç, and Alperen Kara

Subjects
histogram of oriented gradients, image processing, Electronic computers. Computer science, QA75.5-76.95
Abstract

Histograms of oriented gradients (HOG) are still one of the most frequently used low-level features for pattern recognition in images. Despite their great popularity and simple implementation performance of the HOG features almost always has been measured on relatively high quality data which are far from real conditions. To fill this gap we experimentally evaluate their performance in the more realistic conditions, based on images affected by different types of noise, such as Gaussian, quantization, and salt-and-pepper, as well on images distorted by occlusions. Different noise scenarios were tested such anti-distortions during training as well as application of a proper denoising method in the recognition stage. As underpinned with experimental results, the negative impact of distortions and noise on object recognition with HOG features can be significantly reduced by employment of a proper denoising strategy.

Published
2020
Full Text
View/download PDF

27. Expanding the network of highways and expressways in the post-socialist countries of the European Union in the years 2004–2019

Author

Koziarski, Stanisław M.

Subjects
Cities. Urban geography, GF125, Transportation and communications, HE1-9990
Abstract

The aim of this article is to analyze the pace of expansion of the network of highways and expressways over the past 15 years of membership in the European Union. Joining the European Union by such post-socialist countries as: Poland, Lithuania, Latvia, Estonia, the Czech Republic, Slovakia, Hungary, Slovenia (2004), Bulgaria, Romania (2007) and Croatia (2013) as well as the funds received for the development of transport infrastructure significantly contributed to the boom in the pace of road investments in these countries. The existing networks of highways and expressways in the 11 analyzed countries are at the stage of creating major national and international connections. The most advanced expansion of the national highway system is taking place in Poland and Hungary. Stable development, with a slight annual growth of the length of new highways, is taking place in Bulgaria, Croatia, the Czech Republic, Romania, Slovakia and Slovenia. Stagnation in the development of roads of this type is occurring in Lithuania, Latvia and Estonia. Toll roads are operated in 8 out of 11 analyzed countries. The vignette system for highways is used in Bulgaria, the Czech Republic, Slovakia, Slovenia, Romania and Hungary. Toll is paid at tollbooths in Croatia and Poland. Most of the highways in Central Europe are concentrated around the capital agglomerations, e.g. Budapest, Prague, Bratislava, Ljubljana, Zagreb, Sofia and Bucharest, with the exception of Warsaw, which is slowly rebuilding its central position in the network of expressways (A-2, S-7, S-8, S-17) in Poland. The first sections of highways in the discussed countries, except for Poland, were located around the capital cities. The layout of the highway network is often determined by the orography of the terrain, especially the course of mountain ranges, great rivers and coasts. Due to the varied topography, costs are rising, and the construction time of engineering structures such as tunnels, overpasses and bridges is longer, with countries such as Croatia, Slovenia, Slovakia, and recently also the Czech Republic, Poland and Romania being forced to build them.

Published
2020
Full Text
View/download PDF

29. Impact of Low Resolution on Image Recognition with Deep Neural Networks: An Experimental Study

Author

Koziarski Michał and Cyganek Bogusław

Subjects
image recognition, deep neural networks, convolutional neural networks, low resolution, super-resolution, Mathematics, QA1-939, Electronic computers. Computer science, QA75.5-76.95
Abstract

Due to the advances made in recent years, methods based on deep neural networks have been able to achieve a state-of-the-art performance in various computer vision problems. In some tasks, such as image recognition, neural-based approaches have even been able to surpass human performance. However, the benchmarks on which neural networks achieve these impressive results usually consist of fairly high quality data. On the other hand, in practical applications we are often faced with images of low quality, affected by factors such as low resolution, presence of noise or a small dynamic range. It is unclear how resilient deep neural networks are to the presence of such factors. In this paper we experimentally evaluate the impact of low resolution on the classification accuracy of several notable neural architectures of recent years. Furthermore, we examine the possibility of improving neural networks’ performance in the task of low resolution image recognition by applying super-resolution prior to classification. The results of our experiments indicate that contemporary neural architectures remain significantly affected by low image resolution. By applying super-resolution prior to classification we were able to alleviate this issue to a large extent as long as the resolution of the images did not decrease too severely. However, in the case of very low resolution images the classification accuracy remained considerably affected.

Published
2018
Full Text
View/download PDF

30. CCR: A combined cleaning and resampling algorithm for imbalanced data classification

Author

Koziarski Michał and Wożniak Michał

Subjects
machine learning, classification, imbalanced data, preprocessing, oversampling, Mathematics, QA1-939, Electronic computers. Computer science, QA75.5-76.95
Abstract

Imbalanced data classification is one of the most widespread challenges in contemporary pattern recognition. Varying levels of imbalance may be observed in most real datasets, affecting the performance of classification algorithms. Particularly, high levels of imbalance make serious difficulties, often requiring the use of specially designed methods. In such cases the most important issue is often to properly detect minority examples, but at the same time the performance on the majority class cannot be neglected. In this paper we describe a novel resampling technique focused on proper detection of minority examples in a two-class imbalanced data task. The proposed method combines cleaning the decision border around minority objects with guided synthetic oversampling. Results of the conducted experimental study indicate that the proposed algorithm usually outperforms the conventional oversampling approaches, especially when the detection of minority examples is considered.

Published
2017
Full Text
View/download PDF

31. Analysis of group evolution prediction in complex networks.

Author

Stanisław Saganowski, Piotr Bródka, Michał Koziarski, and Przemysław Kazienko

Subjects
Medicine, Science
Abstract

In the world, in which acceptance and the identification with social communities are highly desired, the ability to predict the evolution of groups over time appears to be a vital but very complex research problem. Therefore, we propose a new, adaptable, generic, and multistage method for Group Evolution Prediction (GEP) in complex networks, that facilitates reasoning about the future states of the recently discovered groups. The precise GEP modularity enabled us to carry out extensive and versatile empirical studies on many real-world complex / social networks to analyze the impact of numerous setups and parameters like time window type and size, group detection method, evolution chain length, prediction models, etc. Additionally, many new predictive features reflecting the group state at a given time have been identified and tested. Some other research problems like enriching learning evolution chains with external data have been analyzed as well.

Published
2019
Full Text
View/download PDF

33. How far do they go?: A spatial examination of missing persons from hospitals

Author

Lorna Ferguson and Jacek Koziarski

Subjects
Public Administration, Law, Pathology and Forensic Medicine
Abstract

PurposeMissing person cases are a global issue impacting policing. Among these, those who abscond from hospitals are especially concerning because these reports require collaboration across services, often strain already limited police and hospital resources and present an elevated level of possible harm due to high prevalence of mental illness, disability and/or addiction. Despite this, to-date, there has been a lack of scholarly attention on this phenomenon from a policing perspective. The present study aims to fill this gap by exploring how far missing hospital patients travel and where they are commonly found.Design/methodology/approachUsing a sample of 731 closed case files (2014–2018) from one police service, we identify spatial behaviour patterns specific to this group of missing persons.FindingsResults suggest that most do not leave the hospital grounds or stay within a 5-km radius. Others were found close to the hospital, within city limits and/or returned of their own volition. By identifying these spatial behaviour patterns associated with missing hospital patients, police can refine probable search areas, allocate resources more efficiently, find the missing faster and develop better-informed responses and collaborative policies.Originality/valueOur research represents the first empirical investigation into missing persons from hospital settings through a spatial perspective. Through descriptive statistical and spatial analyses, we determine the distance between the hospital a given individual was reported missing from and the location of where they were ultimately found.

Published
2021

34. Surgical treatment of symptomatic pineal cysts without hydrocephalus—meta-analysis of the published literature

Author

Andreas Unterberg, Steffen Fleck, Martin Májovský, Joham Choque-Velasquez, Andrzej Podgorski, Ali Ansaripour, Charles Teo, Andrzej Koziarski, Per Kristian Eide, Jacky T. Yeung, Moncef Berhouma, Henry W. S. Schroeder, Vladimír Beneš, Angelos G. Kolias, Juha Hernesniemi, Riccardo Masina, Stepan Fedorko, Thomas Santarius, Santarius, Thomas [0000-0002-1416-9566], Apollo - University of Cambridge Repository, Clinicum, HUS Neurocenter, Neurokirurgian yksikkö, and Department of Neurosciences

Subjects
Adult, Male, medicine.medical_specialty, Pediatrics, RESECTION, Nausea, CASE SERIES, Pineal Gland, 3124 Neurology and psychiatry, 03 medical and health sciences, Pineal, 0302 clinical medicine, MANAGEMENT, medicine, Humans, Cyst, Prospective Studies, Symptomatic pineal cyst, 10. No inequality, Prospective cohort study, Retrospective Studies, Non-hydrocephalic symptomatic pineal cyst, Cysts, business.industry, 3112 Neurosciences, Headache, 3126 Surgery, anesthesiology, intensive care, radiology, Review Article - CSF Circulation, medicine.disease, 3. Good health, Hydrocephalus, Treatment Outcome, GLAND, 030220 oncology & carcinogenesis, Meta-analysis, Pineal cyst, Inclusion and exclusion criteria, Female, Surgery, APOPLEXY, Neurology (clinical), Neurosurgery, SUPERIOR COLLICULUS, medicine.symptom, business, 030217 neurology & neurosurgery, Cohort study
Abstract

Background To examine published data and assess evidence relating to safety and efficacy of surgical management of symptomatic pineal cysts without hydrocephalus (nhSPC), we performed a systematic review of the literature and meta-analysis. Methods Following the PRISMA guidelines, we searched Pubmed and SCOPUS for all reports with the query ‘Pineal Cyst’ AND ‘Surgery’ as of March 2021, without constraints on study design, publication year or status (PROSPERO_CRD:42,021,242,517). Assessment of 1537 hits identified 26 reports that met inclusion and exclusion criteria. Results All 26 input studies were either case reports or single-centre retrospective cohorts. The majority of outcome data were derived from routine physician-recorded notes. A total of 294 patients with surgically managed nhSPC were identified. Demographics: Mean age was 29 (range: 4–63) with 77% females. Mean cyst size was 15 mm (5–35). Supracerebellar-infratentorial approach was adopted in 90% of cases, occipital-transtentorial in 9%, and was not reported in 1%. Most patients were managed by cyst resection (96%), and the remainder by fenestration. Mean post-operative follow-up was 35 months (0–228). Presentation: Headache was the commonest symptom (87%), followed by visual (54%), nausea/vomit (34%) and vertigo/dizziness (31%). Other symptoms included focal neurology (25%), sleep disturbance (17%), cognitive impairment (16%), loss of consciousness (11%), gait disturbance (11%), fatigue (10%), ‘psychiatric’ (2%) and seizures (1%). Mean number of symptoms reported at presentation was 3 (0–9). Outcomes: Improvement rate was 93% (to minimise reporting bias only consecutive cases from cohort studies were considered, N = 280) and was independent of presentation. Predictors of better outcomes were large cyst size (OR = 5.76; 95% CI: 1.74–19.02) and resection over fenestration (OR = 12.64; 3.07–52.01). Age predicted worse outcomes (OR = 0.95; 0.91–0.99). Overall complication rate was 17% and this was independent of any patient characteristics. Complications with long-term consequences occurred in 10 cases (3.6%): visual disturbance (3), chronic incisional pain (2), sensory disturbance (1), fatigue (1), cervicalgia (1), cerebellar stroke (1) and mortality due to myocardial infarction (1). Conclusions Although the results support the role of surgery in the management of nhSPCs, they have to be interpreted with a great deal of caution as the current evidence is limited, consisting only of case reports and retrospective surgical series. Inherent to such studies are inhomogeneity and incompleteness of data, selection bias and bias related to assessment of outcome carried out by the treating surgeon in the majority of cases. Prospective studies with patient-reported and objective outcome assessment are needed to provide higher level of evidence.

Published
2021

36. Auxiliary Police Volunteer Experiences and Motivations to Volunteer in Canada

Author

Jacek Koziarski, Christopher D. O’Connor, Tyler Frederick, Victoria Baker, and Kaylee Kosoralo

Subjects
Officer, Service (business), Perceived quality, ComputingMilieux_THECOMPUTINGPROFESSION, Work (electrical), business.industry, Political science, Public relations, business, Law, Variety (cybernetics)
Abstract

Policing has become a shared endeavour among a variety of community stakeholders. Citizens are expected to take on a more active role in securing their own safety. Volunteers are one particular group that has been marshalled to become an essential part of policing. In Canada, volunteers work alongside police officers as auxiliary members and assist in a wide range of activities, such as victim support, safety campaigns, community events, and patrol. Despite auxiliary members actively participating in policing duties, we know little about their experiences or motivations for volunteering. This article presents the results of a survey conducted with auxiliary police personnel at a police service in Canada and discusses their roles and tasks, perceived quality of and ways to improve their experiences, and motivations to volunteer. We conclude by discussing how police services could enhance auxiliary members’ experiences and better integrate this group into regular police officer recruitment efforts.

Published
2021

37. Presenting a uniformed self: Symbols of safety in police auxiliary members' perceptions

Author

Kaylee Kosarolo, Victoria Baker, Tyler Frederick, Jacek Koziarski, and Christopher D. O’Connor

Subjects
Perception, media_common.quotation_subject, Psychology, Law, Social psychology, media_common
Abstract

Volunteers have always played a prominent role in policing. Although known by many names worldwide, auxiliary police in Canada are one particular group of formalized volunteers that have received little research attention. Therefore, through an exploratory survey utilizing both closed and open-ended questions, this article adds to the literature on volunteer police by focusing on how auxiliary members perceived their safety at a police service located in Canada. The findings show how auxiliary members' perceptions of safety were intricately connected to their uniforms, received trainings and associated accoutrements. More specifically, we find that these key elements act as symbols connecting auxiliary members to the extended police family and when they are absent members can feel distanced. Further, safety concerns were expressed as a result of such distancing. The implications of these findings are discussed.

Published
2021

38. The Spatial Concentration, Stability, and Specialization of Mental Health Calls for Service: Evidence in Support of Proactive, Place-Based Interventions

Author

Koziarski, Jacek

Subjects
Mental Health, Proactive, Criminology, Spatial Patterns, Place-Based Interventions, Police
Abstract

For many decades the police have been the de facto responders to persons with perceived mental illness (PwPMI). However, having the police in this role has come with negative repercussions for PwPMI, such as disproportionately experiencing criminalization and use of force. In recognizing these issues, the police—and more recently, the community—have developed responses that either seek to improve interactions between the police and PwPMI or remove the police from this role altogether. However, in either case, these efforts are reactive in nature, responding to crises that arguably could have been prevented had a timelier intervention taken place. Further, evidence on certain police responses to PwPMI, such as Crisis Intervention Teams (CIT) and co-response teams, suggests that they endure deployment-related challenges, thus limiting their reach to PwPMI. Drawing from the Criminology of Place and existing place-based policing strategies, the present dissertation argues that efforts focused on responding to PwPMI should instead be proactively deployed, targeting areas where interactions between police and PwPMI concentrate spatially. Doing so would not only result in efficient deployment of scarce resources but would permit police- and community-based efforts to have a greater reach to PwPMI and thus prevent future interactions with police. To-date, however, there have been few empirical and theoretical investigations into the spatial patterns of PwPMI calls for service that could inform such proactive, place-based efforts. Specifically, we do not currently understand: (1) the degree to which PwPMI calls for service concentrate within certain geographical contexts (such as a small city); (2) whether the degree of PwPMI call concentration and the location of these calls remain stable over time; and (3) what theoretical frameworks explain why PwPMI calls for service occur where they do. Drawing on seven years (2014-2020) of calls for service data from the Barrie Police Service and data from the 2016 Canadian Census, the present dissertation employs various methods of spatial analysis to fills these specific knowledge gaps. Although the theoretical investigation confirmed the findings of previous work that found no association between social disorganization theory and the spatial patterns of PwPMI calls for service, the present dissertation revealed: (1) PwPMI calls for service are highly concentrated within the context of a small city, even more so than what has previously been uncovered in larger jurisdictions; (2) the degree of PwPMI call concentration is stable over time, falling within a narrow proportional bandwidth of spatial units; and (3) PwPMI calls for service, and their concentrations, occur in the same places over time—even during the COVID-19 pandemic—and are thus spatially stable. As such, though more scholarship is needed on theories that might help explain why PwPMI calls occur where they do, the findings of the present dissertation strongly support the proactive, place-based deployment of resources to PwPMI.

Published
2022

40. Radial-Based Oversampling for Multiclass Imbalanced Data Classification

Author

Bartosz Krawczyk, Michal Wozniak, and Michał Koziarski

Subjects
Class (computer programming), Basis (linear algebra), Computer Networks and Communications, Computer science, Process (engineering), business.industry, Novelty, 02 engineering and technology, Machine learning, computer.software_genre, Imbalanced data, Field (computer science), Computer Science Applications, Data set, Artificial Intelligence, 0202 electrical engineering, electronic engineering, information engineering, Oversampling, 020201 artificial intelligence & image processing, Artificial intelligence, business, computer, Software
Abstract

Learning from imbalanced data is among the most popular topics in the contemporary machine learning. However, the vast majority of attention in this field is given to binary problems, while their much more difficult multiclass counterparts are relatively unexplored. Handling data sets with multiple skewed classes poses various challenges and calls for a better understanding of the relationship among classes. In this paper, we propose multiclass radial-based oversampling (MC-RBO), a novel data-sampling algorithm dedicated to multiclass problems. The main novelty of our method lies in using potential functions for generating artificial instances. We take into account information coming from all of the classes, contrary to existing multiclass oversampling approaches that use only minority class characteristics. The process of artificial instance generation is guided by exploring areas where the value of the mutual class distribution is very small. This way, we ensure a smart oversampling procedure that can cope with difficult data distributions and alleviate the shortcomings of existing methods. The usefulness of the MC-RBO algorithm is evaluated on the basis of extensive experimental study and backed-up with a thorough statistical analysis. Obtained results show that by taking into account information coming from all of the classes and conducting a smart oversampling, we can significantly improve the process of learning from multiclass imbalanced data.

Published
2020

41. Expanding the network of highways and expressways in the post-socialist countries of the European Union in the years 2004–2019

Author

Stanisław M. Koziarski

Subjects
Economy, Political science, Post socialist, media_common.cataloged_instance, Cities. Urban geography, GF125, European union, Transportation and communications, HE1-9990, media_common
Abstract

The aim of this article is to analyze the pace of expansion of the network of highways and expressways over the past 15 years of membership in the European Union. Joining the European Union by such post-socialist countries as: Poland, Lithuania, Latvia, Estonia, the Czech Republic, Slovakia, Hungary, Slovenia (2004), Bulgaria, Romania (2007) and Croatia (2013) as well as the funds received for the development of transport infrastructure significantly contributed to the boom in the pace of road investments in these countries. The existing networks of highways and expressways in the 11 analyzed countries are at the stage of creating major national and international connections. The most advanced expansion of the national highway system is taking place in Poland and Hungary. Stable development, with a slight annual growth of the length of new highways, is taking place in Bulgaria, Croatia, the Czech Republic, Romania, Slovakia and Slovenia. Stagnation in the development of roads of this type is occurring in Lithuania, Latvia and Estonia. Toll roads are operated in 8 out of 11 analyzed countries. The vignette system for highways is used in Bulgaria, the Czech Republic, Slovakia, Slovenia, Romania and Hungary. Toll is paid at tollbooths in Croatia and Poland. Most of the highways in Central Europe are concentrated around the capital agglomerations, e.g. Budapest, Prague, Bratislava, Ljubljana, Zagreb, Sofia and Bucharest, with the exception of Warsaw, which is slowly rebuilding its central position in the network of expressways (A-2, S-7, S-8, S-17) in Poland. The first sections of highways in the discussed countries, except for Poland, were located around the capital cities. The layout of the highway network is often determined by the orography of the terrain, especially the course of mountain ranges, great rivers and coasts. Due to the varied topography, costs are rising, and the construction time of engineering structures such as tunnels, overpasses and bridges is longer, with countries such as Croatia, Slovenia, Slovakia, and recently also the Czech Republic, Poland and Romania being forced to build them.

Published
2020

42. A Multiresolution Grid Structure Applied to Seafloor Shape Modeling

Author

Wojciech Maleika, Michał Koziarski, and Paweł Forczmański

Subjects
multiresolution depth grid, data compression, DTM, sea floor, Geography (General), G1-922
Abstract

This paper proposes a method of creating a multiresolution depth grid containing bathymetric data describing a stretch of sea floor. The included literature review presents current solutions in the area of the creation of digital terrain models (DTMs) focusing on methods employing regular grids, with a discussion on the strong and weak points of such an approach. As a basis for the investigations, some important recommendations from the International Hydrographic Organization are provided and are related to the accuracy of created models. The authors propose a novel method of storing DTM data, involving multiresolution depth grids. The paper presents the characteristics of this method, numerical algorithms of a conversion between a regular grid and the multiresolution one, and experiments on typical seafloor surfaces. The results are discussed, focusing on the data reduction rate and the variable resolution of the grid structure. The proposed method can be applied in Geographical Information Systems, especially for the purposes of solving sea survey problems.

Published
2018
Full Text
View/download PDF

44. RB-CCR: Radial-Based Combined Cleaning and Resampling algorithm for imbalanced data classification

Author

Colin Bellinger, Michał Koziarski, and Michal Wozniak

Subjects
FOS: Computer and information sciences, voversampling, Computer Science - Machine Learning, Computer science, Imbalanced data, Machine Learning (cs.LG), Artificial Intelligence, Joint probability distribution, Resampling, Prior probability, Oversampling, business.industry, Pattern recognition, radial basis functions, Class (biology), Statistical classification, ComputingMethodologies_PATTERNRECOGNITION, Asynchronous communication, Binary data, Benchmark (computing), imbalanced data, Artificial intelligence, business, Algorithm, Energy (signal processing), Software
Abstract

In this paper we propose Radial-Based Combined Cleaning and Resampling (RB-CCR) algorithm. RB-CCR utilizes the concept of class potential to refine the energy-based resampling approach of previously proposed CCR algorithm. In particular, RB-CCR exploits the class potential to accurately locate sub-regions of the data-space for synthetic oversampling. The category sub-region for oversampling can be specified as an input parameter to meet domain-specific needs or be automatically selected via cross-validation. The results of the conducted experimental study show that RB-CCR achieves a better precision-recall trade-off than CCR and generally outperforms the state-of-the-art resampling methods in terms of AUC and G-mean., 2021 IEEE 8th International Conference on Data Science and Advanced Analytics (DSAA), October 6-9, 2021, Porto, Portugal

Published
2021

45. CSMOUTE: Combined Synthetic Oversampling and Undersampling Technique for Imbalanced Data Classification

Author

Michał Koziarski

Subjects
FOS: Computer and information sciences, Computer Science - Machine Learning, Artificial neural network, Computer science, business.industry, Machine Learning (stat.ML), Pattern recognition, Imbalanced data, Machine Learning (cs.LG), Support vector machine, Statistical classification, Statistics - Machine Learning, Undersampling, Outlier, Oversampling, Artificial intelligence, business, Interpolation
Abstract

In this paper we propose a novel data-level algorithm for handling data imbalance in the classification task, Synthetic Majority Undersampling Technique (SMUTE). SMUTE leverages the concept of interpolation of nearby instances, previously introduced in the oversampling setting in SMOTE. Furthermore, we combine both in the Combined Synthetic Oversampling and Undersampling Technique (CSMOUTE), which integrates SMOTE oversampling with SMUTE undersampling. The results of the conducted experimental study demonstrate the usefulness of both the SMUTE and the CSMOUTE algorithms, especially when combined with more complex classifiers, namely MLP and SVM, and when applied on datasets consisting of a large number of outliers. This leads us to a conclusion that the proposed approach shows promise for further extensions accommodating local data characteristics, a direction discussed in more detail in the paper.

Published
2021

46. DiagSet: a dataset for prostate cancer histopathological image classification

Author

Koziarski, Micha��, Cyganek, Bogus��aw, Olborski, Bogus��aw, Antosz, Zbigniew, ��ydak, Marcin, Kwolek, Bogdan, W��sowicz, Pawe��, Buka��a, Andrzej, Swad��ba, Jakub, and Sitkowski, Piotr

Subjects
FOS: Computer and information sciences, Computer Science - Machine Learning, Computer Vision and Pattern Recognition (cs.CV), Image and Video Processing (eess.IV), FOS: Electrical engineering, electronic engineering, information engineering, Computer Science - Computer Vision and Pattern Recognition, Electrical Engineering and Systems Science - Image and Video Processing, Machine Learning (cs.LG)
Abstract

Cancer diseases constitute one of the most significant societal challenges. In this paper we introduce a novel histopathological dataset for prostate cancer detection. The proposed dataset, consisting of over 2.6 million tissue patches extracted from 430 fully annotated scans, 4675 scans with assigned binary diagnosis, and 46 scans with diagnosis given independently by a group of histopathologists, can be found at https://ai-econsilio.diag.pl. Furthermore, we propose a machine learning framework for detection of cancerous tissue regions and prediction of scan-level diagnosis, utilizing thresholding and statistical analysis to abstain from the decision in uncertain cases. During the experimental evaluation we identify several factors negatively affecting the performance of considered models, such as presence of label noise, data imbalance, and quantity of data, that can serve as a basis for further research. The proposed approach, composed of ensembles of deep neural networks operating on the histopathological scans at different scales, achieves 94.6% accuracy in patch-level recognition, and is compared in a scan-level diagnosis with 9 human histopathologists.

Published
2021

47. Double pituitary adenomas in a large surgical series

Author

Andrzej Koziarski, Grzegorz Zieliński, Maria Maksymowicz, Monika Pękul, and Emir Sajjad

Subjects
Adenoma, Adult, Male, medicine.medical_specialty, Transsphenoidal approach, Endocrinology, Diabetes and Metabolism, medicine.medical_treatment, Neurosurgery, Article, Young Adult, Endocrinology, Pituitary adenoma, Double pituitary adenoma, Acromegaly, medicine, Humans, Pituitary Neoplasms, Pituitary ACTH Hypersecretion, Carney complex, Aged, Retrospective Studies, Transsphenoidal surgery, Surgical treatment, business.industry, Medical record, Incidence (epidemiology), Pituitary tumors, Middle Aged, medicine.disease, Immunohistochemistry, Magnetic Resonance Imaging, Female, Radiology, business
Abstract

Purpose To explore the incidence of double pituitary adenomas in a tertiary center for pituitary surgery and asses their clinical, imaging and histopathological features. Methods The medical records of the patients operated on for pituitary tumors at the Department of Neurosurgery of Military Institute of Medicine in Warsaw, Poland between the years 2003 and 2018 were retrospectively analyzed. Among the 3270 treated patients, the diagnosis of double pituitary adenoma was established in 22 patients. Clinical, laboratory, detailed histopathological and diagnostics imaging data were collected and analyzed. Results There were 21 cases of synchronous and one case of asynchronous double pituitary adenoma. The main clinical finding was acromegaly (12/22) followed by Cushing’s disease (3/22). The diagnosis of synchronous double pituitary adenoma was suspected in the preoperative MRI in 11 patients. In the remaining patients the diagnosis of contiguous double pituitary adenoma was confirmed in the histopathological examination. There was no predilection for gender and the mean observation time was 74.2 months. In one case of Cushing’s disease the occurrence of double pituitary adenoma led to the initial failure of achieving hormonal remission. One patient presented with double pituitary adenomas as a manifestation of Carney complex. Conclusions Double pituitary adenoma is a rare entity that can pose a significant challenge especially in the setting of Cushing’s disease. Careful inspection of preoperative MRI and diagnostic work-up before transsphenoidal surgery and thorough histopathological microscopic examinations with immunohistochemical staining for all pituitary hormones is essential for establishing the diagnosis of double pituitary adenoma.

Published
2019

48. Are type I dural arteriovenous fistulas safe? Single-centre experience of endovascular treatment of dural arteriovenous fistulas

Author

Jerzy Narloch, Piotr Piasecki, Andrzej Koziarski, Piotr Zięcina, and Krzysztof Brzozowski

Subjects
medicine.medical_specialty, Original Paper, business.industry, Incidence (epidemiology), 05 social sciences, Arteriovenous fistula, 050801 communication & media studies, medicine.disease, AVF, Asymptomatic, Surgery, Onyx, Single centre, 0508 media and communications, Contrast used, Dural arteriovenous fistulas, medicine, interventional neuroradiology, embolisation, haemorrhage, Endovascular treatment, medicine.symptom, business, Interventional neuroradiology
Abstract

Introduction: There are mixed reports on the incidence of intracranial haemorrhage in patients with dural arteriovenous fistulas. We assessed new proposed risk factors (i.a. number of outflows and outflow diameter) of intracerebral haemorrhage due to intracranial dural arteriovenous fistula and presented our personal experience in endovascular treatment of dural arteriovenous fistulas. Material and methods: The patient database from January 2006 and December 2016 was reviewed, and 25 patients with 28 dural arteriovenous fistulas were identified. Results: 50% of patients presented with intracerebral haemorrhage. Multiple dural fistulas occurred in 12% of patients. Spearman's rank correlation coefficient revealed that there was a strong association between Cognard classification type and time needed to treat (r = 0.59, p < 0.05), as well as the volume of contrast used (r = 0.77, p < 0.05). Infratentorial (r = 0.53, p < 0.05) and right-sided (r = 0.66, p < 0.05) localisation were more challenging to treat. Bleeding was associated with poorer clinical outcome (r = 0.48, p < 0.05). No significant differences were found between the non-haemorrhagic group and the haemorrhagic group regarding the number of outflows (p = 0.459) and largest outflow diameter (p = 0.298). Clinical evaluation at follow-up was as follows: 56% of patients were asymptomatic, 24% had non-significant disability, maintaining independency, 16% had moderate disability, and 8% died - one in the course of intracerebral haemorrhage and one due to other sustained injuries. There were no reported embolisation- related complications. Conclusions: To conclude, regardless of presentation, both symptomatic and asymptomatic dural arteriovenous fistulas deserve clinical attention, structured evaluation, and follow-up. Type I fistulas were associated with haemorrhage in 1/3 of all cases. Overall our results indicate that the risk of haemorrhage and dire consequences is multifactorial.

Published
2019

49. Factors Associated with Recurrence of Varicose Veins after Thermal Ablation: Results of The Recurrent Veins after Thermal Ablation Study

Author

R. G. Bush, P. Bush, J. Flanagan, R. Fritz, T. Gueldner, J. Koziarski, K. McMullen, and G. Zumbro

Subjects
Technology, Medicine, Science
Abstract

Background. The goal of this retrospective cohort study (REVATA) was to determine the site, source, and contributory factors of varicose vein recurrence after radiofrequency (RF) and laser ablation. Methods. Seven centers enrolled patients into the study over a 1-year period. All patients underwent previous thermal ablation of the great saphenous vein (GSV), small saphenous vein (SSV), or anterior accessory great saphenous vein (AAGSV). From a specific designed study tool, the etiology of recurrence was identified. Results. 2,380 patients were evaluated during this time frame. A total of 164 patients had varicose vein recurrence at a median of 3 years. GSV ablation was the initial treatment in 159 patients (RF: 33, laser: 126, 52 of these patients had either SSV or AAGSV ablation concurrently). Total or partial GSV recanalization occurred in 47 patients. New AAGSV reflux occurred in 40 patients, and new SSV reflux occurred in 24 patients. Perforator pathology was present in 64% of patients. Conclusion. Recurrence of varicose veins occurred at a median of 3 years after procedure. The four most important factors associated with recurrent veins included perforating veins, recanalized GSV, new AAGSV reflux, and new SSV reflux in decreasing frequency. Patients who underwent RF treatment had a statistically higher rate of recanalization than those treated with laser.

Published
2014
Full Text
View/download PDF

50. Impact of Low Resolution on Image Recognition with Deep Neural Networks: An Experimental Study

Author

Bogusław Cyganek and Michał Koziarski

Subjects
021110 strategic, defence & security studies, Computer science, business.industry, Applied Mathematics, Low resolution, 0211 other engineering and technologies, ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION, super-resolution, 02 engineering and technology, QA75.5-76.95, image recognition, deep neural networks, Electronic computers. Computer science, convolutional neural networks, 0202 electrical engineering, electronic engineering, information engineering, Computer Science (miscellaneous), QA1-939, Deep neural networks, 020201 artificial intelligence & image processing, Computer vision, Artificial intelligence, business, Engineering (miscellaneous), Mathematics, low resolution
Abstract

Due to the advances made in recent years, methods based on deep neural networks have been able to achieve a state-of-the-art performance in various computer vision problems. In some tasks, such as image recognition, neural-based approaches have even been able to surpass human performance. However, the benchmarks on which neural networks achieve these impressive results usually consist of fairly high quality data. On the other hand, in practical applications we are often faced with images of low quality, affected by factors such as low resolution, presence of noise or a small dynamic range. It is unclear how resilient deep neural networks are to the presence of such factors. In this paper we experimentally evaluate the impact of low resolution on the classification accuracy of several notable neural architectures of recent years. Furthermore, we examine the possibility of improving neural networks’ performance in the task of low resolution image recognition by applying super-resolution prior to classification. The results of our experiments indicate that contemporary neural architectures remain significantly affected by low image resolution. By applying super-resolution prior to classification we were able to alleviate this issue to a large extent as long as the resolution of the images did not decrease too severely. However, in the case of very low resolution images the classification accuracy remained considerably affected.

Published
2018

Catalog

Books, media, physical & digital resources
See catalog results