1. |
- Karlsson, Isak, 1987-
(author)
-
Order in the random forest
- 2017
-
Doctoral thesis (other academic/artistic)abstract
- In many domains, repeated measurements are systematically collected to obtain the characteristics of objects or situations that evolve over time or other logical orderings. Although the classification of such data series shares many similarities with traditional multidimensional classification, inducing accurate machine learning models using traditional algorithms are typically infeasible since the order of the values must be considered.In this thesis, the challenges related to inducing predictive models from data series using a class of algorithms known as random forests are studied for the purpose of efficiently and effectively classifying (i) univariate, (ii) multivariate and (iii) heterogeneous data series either directly in their sequential form or indirectly as transformed to sparse and high-dimensional representations. In the thesis, methods are developed to address the challenges of (a) handling sparse and high-dimensional data, (b) data series classification and (c) early time series classification using random forests. The proposed algorithms are empirically evaluated in large-scale experiments and practically evaluated in the context of detecting adverse drug events.In the first part of the thesis, it is demonstrated that minor modifications to the random forest algorithm and the use of a random projection technique can improve the effectiveness of random forests when faced with discrete data series projected to sparse and high-dimensional representations. In the second part of the thesis, an algorithm for inducing random forests directly from univariate, multivariate and heterogeneous data series using phase-independent patterns is introduced and shown to be highly effective in terms of both computational and predictive performance. Then, leveraging the notion of phase-independent patterns, the random forest is extended to allow for early classification of time series and is shown to perform favorably when compared to alternatives. The conclusions of the thesis not only reaffirm the empirical effectiveness of random forests for traditional multidimensional data but also indicate that the random forest framework can, with success, be extended to sequential data representations.
|
|
2. |
- Seçilmiş, Deniz, 1991-
(author)
-
Improving the accuracy of gene regulatory network inference from noisy data
- 2021
-
Doctoral thesis (other academic/artistic)abstract
- Gene regulatory networks (GRNs) control physiological and pathological processes in a living organism, and their accurate inference from measured gene expression can identify therapeutic mechanisms for complex diseases such as cancers. The biggest obstacle in achieving the accurate reconstruction of GRNs is called ‘noise’, which considerably alters the measured gene expression because the noise generally dominates the biological signal. This situation needs to be addressed carefully so that GRN inference methods do not estimate a fit to the noise instead of the underlying biological signal. Potential noise compensation approaches are a must if the goal is to reconstruct the true system. To this end, within the scope of this doctoral thesis, I developed two methods that, in different ways, overcome the obstacles introduced by noise in gene expression data. Method 1 allows the collection of more informative subsets of genes whose expression is not as highly affected as those which cause the system to be overall uninformative. Method 2 infers a perturbation design that is better suited to the gene expression data than the originally intended design, and therefore produces more accurate GRNs at high noise levels. Furthermore, a benchmark study was carried out which compares the methodological backgrounds of GRN inference methods in terms of whether they utilize knowledge of the perturbation design or not, which clearly shows that utilization of the perturbation design is essential for accurate inference of GRNs. Finally a method is presented to improve GRN inference accuracy by selecting the GRN with the optimal sparsity based on information theoretical criteria. The three new methods (PAPERS I, II and IV) can also be used together, which is shown in this thesis to improve the GRN inference accuracy considerably more than the methods separately. As inference of accurate GRNs is a major challenge in gene regulation, the methods presented in this thesis represent an important contribution to move the field forward.
|
|