By adaptation of our explanation methodology we show that classifications of PRBF (Probabilistic Radial Base Function network), which is an efficient black box classifier, can be efffectively explained. We demonstrate this on artificial and real-world medical domains.
COBISS.SI-ID: 9365588
The paper describes the use of expert’s knowledge in practice and the efficiency of a recently developed technique called argument-based machine learning (ABML) in the knowledge elicitation process. ABML guides the expert to explain critical special cases which cannot be handled automatically by machine learning. This very efficiently reduces the expert’s workload, and combines it with automatically learned knowledge. We are developing a neurological decision support system to help the neurologists differentiate between three types of tremors: Parkinsonian, essential, and mixed tremor (comorbidity). The system is intended to act as a second opinion for the neurologists, and most importantly to help them reduce the number of patients in the “gray area” that require a very costly further examination (DaTSCAN).
COBISS.SI-ID: 30199257
We compared the ability of CLIP and iCLIP methods to identify UV-induced RNA sites interacting with proteins. The comparison was done on six published and two new experimental data sets. Over 80% of iCLIP reads truncated at the cross-link site, while only 10% of CLIP reads included a deletion, which is an indication for cross-linking. We have demonstrated that iCLIP provides a higher positional precision and is able to identify more RNA sites interacting with the given protein.
COBISS.SI-ID: 9319764
We report on Nimfa, an open-source Python library that provides a unified interface to nonnegative matrix factorization algorithms. It includes implementations of state-of-the-art factorization methods, initialization approaches, and quality scoring. Both dense and sparse matrix representations are supported.
COBISS.SI-ID: 9067604
Based on microarray data we identified potential biomarkers for Alzheimer’s disease using three feature selection methods: information gain, mean decrease accuracy of random forest and a wrapper of genetic algorithm and support vector machine (GA/SVM). Information gain and random forest are two commonly used methods. GA/SVM is rarely used for the analysis of microarray data, but we show that it is able to identify genes capable of classifying tissues into different classes at least as well as the two reference methods. Compared to the other methods, GA/SVM has the advantage of finding small, less redundant sets of genes that, in combination, show superior classification characteristics.
COBISS.SI-ID: 26200615