WWWitold Wydmański

Witold Wydmański

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Witold Wydmański has a total of 32 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • LLM-as-a-judge is bad, based on AI attempting the exam qualifying for the member of the Polish National Board of Appeal

    by Michał Karp, Anna Kubaszewska, Magdalena Król, Robert Król, Witold Wydmański, Aleksander Smywiński-Pohl, Mateusz Szymański


  • Machine Unlearning for Recommendation Systems: An Insight

    by Bhavika Sachdeva, Harshita Rathee, Sristi, Arun Sharma, Witold Wydmański


  • Open Access

    Machine Unlearning for Recommendation Systems: An Insight

    by Bhavika Sachdeva, Harshita Rathee, Sristi, Arun Sharma, Witold Wydmański

    This review explores machine unlearning (MUL) in recommendation systems, addressing adaptability, personalization, privacy, and bias challenges. Unlike traditional models, MUL dynamically adjusts system knowledge based on shifts in user preferences and ethical considerations. The paper critically examines MUL's basics, real-world applications, and challenges like algorithmic transparency. It sifts through literature, offering insights into how MUL could transform recommendations, discussing user trust, and suggesting paths for future research in responsible and user-focused artificial intelligence (AI). The document guides researchers through challenges involving the trade-off between personalization and privacy, encouraging contributions to meet practical demands for targeted data removal. Emphasizing MUL's role in secure and adaptive machine learning, the paper proposes ways to push its boundaries. The novelty of this paper lies in its exploration of the limitations of the methods, which highlights exciting prospects for advancing the field.

  • Open Access

    HyperTab: Hypernetwork Approach for Deep Learning on Small Tabular Datasets

    by Witold Wydmański, Oleksii Bulenok, Marek Śmieja

    Deep learning has achieved impressive performance in many domains, such as computer vision and natural language processing, but its advantage over classical shallow methods on tabular datasets remains questionable. It is especially challenging to surpass the performance of tree-like ensembles, such as XGBoost or Random Forests, on small-sized datasets (less than 1k samples). To tackle this challenge, we introduce HyperTab, a hypernetwork-based approach to solving small sample problems on tabular datasets. By combining the advantages of Random Forests and neural networks, HyperTab generates an ensemble of neural networks, where each target model is specialized to process a specific lower-dimensional view of the data. Since each view plays the role of data augmentation, we virtually increase the number of training samples while keeping the number of trainable parameters unchanged, which prevents model overfitting. We evaluated HyperTab on more than 40 tabular datasets of a varying number of samples and domains of origin, and compared its performance with shallow and deep learning models representing the current state-of-the-art. We show that HyperTab consistently outranks other methods on small data (with a statistically significant difference) and scores comparable to them on larger datasets. We make a python package with the code available to download at https://pypi.org/project/hypertab/


  • Open Access

    SARS-CoV-2 genome diversity at the binding sites of oligonucleotides used for COVID-19 diagnosis

    by Renan Valieris, Michał B. Kowalski, Alina Frolova, Witold Wydmański, Johnathan Foox, Giovana T. Torrezan, Ewelina Pośpiech, Wojciech Branicki, Kasthuri Venkateswaran, Bharath Prithiviraj, et al.

    Importance: SARS-CoV-2 genomic variants impacts the overall sensitivity of COVID-19 diagnosis, leading to false-negative diagnosis and the continued spread of the virus. Objective: To evaluate how nucleotide variability in target primer binding sites of the SARS-CoV-2 genomes may impact diagnosis using different recommended primer/probe sets, as well as to suggest the best primer/probes for diagnosis. Design: We downloaded 105,118 public SARS-CoV-2 genomes from GISAID (Sept, 25th, 2020), removed genomes of apparent worst quality (genome length 5% ambiguous bases) and missing metadata, and performed an analysis of complementarity for the 13 most used diagnostic primers/probe sets for RT-PCR detection. We calculated the N rate and % of genome recovery, with all primer/probe-sets considering viral origin and clade. Results: Our findings indicate that currently, the Paris_nCoV-IP2, -IP4 and WHO|E_Sarbeco primer/probe sets for COVID-19, to perform the best diagnostically worldwide, recovering >99.5% of the good quality SARS-CoV-2 genomes from GISAID, with no mismatches. The Chinese_CDC|2019-nCoV-NP primer/probe set, among the first to be designed during the pandemic, was the most susceptible to currently most abundant SARS-CoV-2 variants. Mismatches encompassing the binding sites for this set are more frequent in Clade-GR and are highly prevalent in over 30 countries globally, including Brazil and India, two of the hardest hit countries. Conclusions: Detection of SARS-CoV-2 in patients may be hampered by significant variability in parts of the viral genome that are targeted by some widely used primer sets. The geographic distribution of different viral clades indicates that continuous assessment of primer sets via sequencing-based surveillance and viral evolutionary analysis is critical to accurate diagnostics. This study highlights sequence variance in target regions that may reduce the efficiency of primer:target hybridization that in turn may lead to the undetected spread of the virus. As such, due to this variance, the Chinese_CDC|2019-nCoV-NP-set should be used with caution, or avoided, especially in countries with high prevalence of the GR clade. Key Points: Question: How variable are the binding-sites of primers/probes used for COVID-19 diagnosis? Findings: We investigated nucleotide variations in primer-binding sites used for COVID-19 diagnosis, in 93,143 SARS-CoV-2 genomes, and found primer sets targeting regions of increasingly nucleotide variance over time, such as the Chinese_CDC|2019-nCoV-NP. The frequency of these variations is higher in Clade-GR whose frequency is increasing worldwide. Paris_nCoV-IP2, IP4 and WHO|E_Sarbeco performed best. Meaning: We suggest the use of some sets to be halted and reinforce the importance of a continuous surveillance of SARS-CoV-2 variations to prompt the use of the best primers.