| Popis: |
Support vector machine (SVM) based machine learning is used in a wide range of domains. It represents a family of supervised machine learning algorithms and is most commonly used for binary classification tasks. It can also be extended to multi-label problems which are specializations of multi-task classification. We use an early stage SVM implementation, called PermonSVM, to implement a one versus all multi-label method to classify and predict protein-compound activities in chemogenomics. The white paper highlights the VI-HPS tools Score-P, Cube and Vampir, as used during the early development and improvement processes of PermonSVM. We apply those tools to identify and analyze a bottleneck in the early PermonSVM implementation, and verify its final iteration. |
