No description available.

No description available.

No description available.

The abundance of next-generation sequencing (NGS) data has encouraged the adoption of machine learning methods to aid in the diagnosis and treatment of human disease. In particular, the last decade has shown the extensive use of predictive analytics in cancer research due to the prevalence of rich cellular descriptions of genetic and transcriptomic profiles of cancer cells. Despite the availability of wide-ranging forms of genomic data, few predictive models are designed to leverage multidimensional data sources. In this paper, we introduce a deep learning approach using neural network based information fusion to facilitate the integration of multi-platform genomic data, and the prediction of cancer cell sub-class. We propose the dGMU (deep gated multimodal unit), a series of multiplicative gates that can learn intermediate representations between multi-platform genomic data and improve cancer cell stratification. We also provide a framework for interpretable dimensionality reduction and assess several methods that visualize and explain the decisions of the underlying model. Experimental results on nine cancer types and four forms of NGS data (copy number variation, simple nucleotide variation, RNA expression, and miRNA expression) showed that the dGMU model improved the classification agreement of unimodal approaches and outperformed other fusion strategies in class accuracy. The results indicate that deep learning architectures based on multiplicative gates have the potential to expedite representation learning and knowledge integration in the study of cancer pathogenesis. / Thesis / Master of Science (MSc)

Andromeda is an interactive visualization tool that projects high-dimensional data into a scatterplot-like visualization using Weighted Multidimensional Scaling (WMDS). The visualization can be explored through surface-level interaction (viewing data values), parametric interaction (altering underlying parameterizations), and observation-level interaction (directly interacting with projected points). This thesis presents analyses on the collaborative utility of Andromeda in a middle school class and the insights college-level students generate when using Andromeda. The first study discusses how a middle school class collaboratively used Andromeda to explore and compare their engineering designs. The students analyzed their designs, represented as high-dimensional data, as a class. This study shows promise for introducing collaborative data analysis to middle school students in conjunction with other technical concepts such as the engineering design process. Participants in the study on college-level students were given a version of Andromeda, with access to different interactions, and were asked to generate insights on a dataset. By applying a novel visualization evaluation methodology on students' natural language insights, the results of this study indicate that students use different vocabulary supported by the interactions available to them, but not equally. The implications, as well as limitations, of these two studies are further discussed. / Master of Science / Data is often high-dimensional. A good example of this is a spreadsheet with many columns. Visualizing high-dimensional data is a difficult task because it must capture all information in 2 or 3 dimensions. Andromeda is a tool that can project high-dimensional data into a scatterplot-like visualization. Data points that are considered similar are plotted near each other and vice versa. Users can alter how important certain parts of the data are to the plotting algorithm as well as move points directly to update the display based on the user-specified layout. These interactions within Andromeda allow data analysts to explore high-dimensional data based on their personal sensemaking processes. As high dimensional thinking and exploratory data analysis are being introduced into more classrooms, it is important to understand the ways in which students analyze high-dimensional data. To address this, this thesis presents two studies. The first study discusses how a middle school class used Andromeda for their engineering design assignments. The results indicate that using Andromeda in a collaborative way enriched the students' learning experience. The second study analyzes how college-level students, when given access to different interaction types in Andromeda, generate insights into a dataset. Students use different vocabulary supported by the interactions available to them, but not equally. The implications, as well as limitations, of these two studies are further discussed.

2025 / Function approximation is a fundamental aspect of computational models and machine learning, often relying on neural networks due to their ability to effectively model complex functions and relationships. However, neural networks can be computationally intensive and lack interpretability. In this thesis, we explore an alternative approach to approximating two-dimensional (2D) functions by decomposing them into multiple one-dimensional (1D) approximations. Our method aims to enhance computational efficiency and interpretability while maintaining high approximation quality. We propose a framework that projects to approximate 2D functions through a series of 1D interpolations and also uses greedy sampling. By generating uniformly distributed projections and projecting pixel coordinates onto these projections, we form 1D curves and use interpolation to predict the values of the original function. Linear interpolation is employed for its simplicity and speed in estimating values between sampled points. A greedy algorithm is used to select sampling points that significantly reduce approximation error, optimizing the sampling strategy. We conducted extensive experiments on some images to evaluate the performance of our method. Metrics such as Mean Squared Error (MSE) and Peak Signal-to-Noise Ratio (PSNR) were used to assess reconstruction quality. Additionally, we ran neural network model and some other traditional models for comparison. Our results demonstrate that the proposed method provides a different focus compared to other methods, especially excelling in the restoration of high-contrast details in images. The findings suggest that multiple 1D approximations can reconstruct 2D functions with efficiency. Contrary to our initial intuition, the results reveal that increasing the number of sample points has a more significant impact on reconstruction quality than increasing the number of projections. Specifically, we observed that under the same parameter count, using as many sample points as possible led to better reconstruction results. Increasing the number of projections, while beneficial for reducing artifacts, has a less pronounced effect compared to increasing sample points. However, adding more projections can improve edge clarity and enhance the accuracy of each step in the greedy selection process, which helps in achieving better sample point locations during reconstruction. Additionally, we tested various sampling methods, such as uniform sampling and greedy MSE selection, and found that greedy selection of sample points based on MSE yielded significantly improved clarity, particularly around key features of the image. The experiments also showed that incorporating spatial diversity and edge information into the selection process did not always yield better results, highlighting the importance of selecting sample points that balance both edge and surrounding details. This work contributes to the field by providing an alternative method for function approximation that addresses some limitations of neural networks, particularly in terms of computational efficiency. Future work includes extending the approach to higher-dimensional data, exploring advanced interpolation techniques, and integrating the method with machine learning models to balance performance and transparency. Additionally, further research is needed to optimize the balance between projections and sample points to achieve the best reconstruction quality under different parameter constraints.

Research on Machine learning (ML) explainability has received a lot of focus in recent times. The interest, however, mostly focused on supervised models, while other ML fields have not had the same level of attention. Despite its usefulness in a variety of different fields, unsupervised learning explainability is still an open issue. In this paper, we present a Visual Analytics framework based on eXplainable AI (XAI) methods to support the interpretation of Dimensionality reduction methods. The framework provides the user with an interactive and iterative process to investigate and explain user-perceived patterns for a variety of DR methods by using XAI methods to explain a supervised method trained on the selected data. To evaluate the effectiveness of the proposed solution, we focus on two main aspects: the quality of the visualization and the quality of the explanation. This challenge is tackled using both quantitative and qualitative methods, and due to the lack of pre-existing test data, a new benchmark has been created. The quality of the visualization is established using a well-known survey-based methodology, while the quality of the explanation is evaluated using both case studies and a controlled experiment, where the generated explanation accuracy is evaluated on the proposed benchmark. The results show a strong capacity of our framework to generate accurate explanations, with an accuracy of 89% over the controlled experiment. The explanation generated for the two case studies yielded very similar results when compared with pre-existing, well-known literature on ground truths. Finally, the user experiment generated high quality overall scores for all assessed aspects of the visualization.

This PhD work constitutes a series of inter-disciplinary studies that use biologically plausible computational techniques and experiments with human subjects in analyzing facial expressions. The performance of the computational models and human subjects in terms of accuracy and response time are analyzed. The computational models process images in three stages. This includes: Preprocessing, dimensionality reduction and Classification. The pre-processing of face expression images includes feature extraction and dimensionality reduction. Gabor filters are used for feature extraction as they are closest biologically plausible computational method. Various dimensionality reduction methods: Principal Component Analysis (PCA), Curvilinear Component Analysis (CCA) and Fisher Linear Discriminant (FLD) are used followed by the classification by Support Vector Machines (SVM) and Linear Discriminant Analysis (LDA). Six basic prototypical facial expressions that are universally accepted are used for the analysis. They are: angry, happy, fear, sad, surprise and disgust. The performance of the computational models in classifying each expression category is compared with that of the human subjects. The Effect size and Encoding face enable the discrimination of the areas of the face specific for a particular expression. The Effect size in particular emphasizes the areas of the face that are involved during the production of an expression. This concept of using Effect size on faces has not been reported previously in the literature and has shown very interesting results. The detailed PCA analysis showed the significant PCA components specific for each of the six basic prototypical expressions. An important observation from this analysis was that with Gabor filtering followed by non linear CCA for dimensionality reduction, the dataset vector size may be reduced to a very small number, in most cases it was just 5 components. The hypothesis that the average response time (RT) for the human subjects in classifying the different expressions is analogous to the distance measure of the data points from the classification hyper-plane was verified. This means the harder a facial expression is to classify by human subjects, the closer to the classifying hyper-plane of the classifier it is. A bi-variate correlation analysis of the distance measure and the average RT suggested a significant anti-correlation. The signal detection theory (SDT) or the d-prime determined how well the model or the human subjects were in making the classification of an expressive face from a neutral one. On comparison, human subjects are better in classifying surprise, disgust, fear, and sad expressions. The RAW computational model is better able to distinguish angry and happy expressions. To summarize, there seems to some similarities between the computational models and human subjects in the classification process.

This thesis centers around dimensionality reduction and its usage on landmark-type data which are often used in anthropology and morphometrics. In particular we focus on non-linear dimensionality reduction methods - locally linear embedding and multidimensional scaling. We introduce a new approach to dimensionality reduction called multipass dimensionality reduction and show that improves the quality of classification as well as requiring less dimensions for successful classification than the traditional singlepass methods.

Random projection method is one of the important tools for the dimensionality reduction of data which can be made efficient with strong error guarantees. In this thesis, we focus on linear transforms of high dimensional data to the low dimensional space satisfying the Johnson-Lindenstrauss lemma. In addition, we also prove some theoretical results relating to the projections that are of interest when applying them in practical applications. We show how the technique can be applied to synthetic data with probabilistic guarantee on the pairwise distance. The connection between dimensionality reduction and compressed sensing is also discussed.