Emilia Oikarinen

Bachelor's theses are a central component of undergraduate science education, but they also pose challenges for large-scale supervision under limited instructional resources. To address this issue, we piloted a model allowing students to write their theses in pairs, an approach common in universities of applied sciences but rare in research-intensive universities. This discussion paper reports on the design, implementation and evaluation of a pilot conducted in autumn 2024 at the University of Helsinki. The findings suggest that co-authored theses may reduce supervisory workload without compromising academic quality. Student feedback highlighted increased engagement, peer support and support for self-regulation, although challenges such as unequal division of labour remained. Supervisors reported reduced cognitive load and supported the model's future use. Based on the pilot, co-authored theses are now an official option in the programme. We conclude by discussing pedagogical implications, assessment practices, and recommendations for supporting collaborative thesis-writing processes.

Learning and teaching at scale became especially topical and important matters due to the COVID-19 pandemic. The obvious way to that end is to use Massive Open Online Courses (MOOCs). However, the pedagogics and student behaviour between online and classroom implementations are far from similar, and successful implementation of a MOOC requires an understanding of this. Typically, the dropout rates of MOOCs are pretty high. Most of the dropouts happen in the very early stages of the course. This paper studies how and when students commit to MOOC courses. For this study, we have selected four courses with different implementations from our extensive portfolio of MOOC courses at University of Helsinki. Based on the results, the students who complete the first two parts (or weeks) of a more prolonged course are usually quite committed to completing the entire course. The behaviour varies slightly between scheduled and unscheduled courses and between courses of different topics. The inner and outer motivations also play a role in commitment. Regardless, it seems that the authors of MOOCs should pay special attention to making the first steps as engaging as possible.

A fundamental problem in visual data exploration concerns whether observed patterns are true or merely random noise. This problem is especially pertinent in visual analytics, where the user is presented with a barrage of patterns, without any guarantees of their statistical validity. Recently this problem has been formulated in terms of statistical testing and the multiple comparisons problem. In this paper, we identify two levels of multiple comparisons problems in visualization: the within-view and the between-view problem. We develop a statistical testing procedure for interactive data exploration that controls the family-wise error rate on both levels. The procedure enables the user to determine the compatibility of their assumptions about the data with visually observed patterns. We present use-cases where we visualize and evaluate patterns in real-world data.

Real-world datasets are often characterised by outliers; data items that do not follow the same structure as the rest of the data. These outliers might negatively influence modelling of the data. In data analysis it is, therefore, important to consider methods that are robust to outliers. In this paper we develop a robust regression method that finds the largest subset of data items that can be approximated using a sparse linear model to a given precision. We show that this can yield the best possible robustness to outliers. However, this problem is NP-hard and to solve it we present an efficient approximation algorithm, termed SLISE. Our method extends existing state-of-the-art robust regression methods, especially in terms of speed on high-dimensional datasets. We demonstrate our method by applying it to both synthetic and real-world regression problems.

Running large-eddy simulations (LESs) can be burdensome and computationally too expensive from the application point of view, for example, to support urban planning. In this study, regression models are used to replicate modelled air pollutant concentrations from LES in urban boulevards. We study the performance of regression models and discuss how to detect situations where the models are applied outside their training domain and their outputs cannot be trusted. Regression models from 10 different model families are trained and a cross-validation methodology is used to evaluate their performance and to find the best set of features needed to reproduce the LES outputs. We also test the regression models on an independent testing dataset. Our results suggest that in general, log-linear regression gives the best and most robust performance on new independent data. It clearly outperforms the dummy model which would predict constant concentrations for all locations (multiplicative minimum RMSE (mRMSE) of 0.76 vs. 1.78 of the dummy model). Furthermore, we demonstrate that it is possible to detect concept drift, i.e. situations where the model is applied outside its training domain and a new LES run may be necessary to obtain reliable results. Regression models can be used to replace LES simulations in estimating air pollutant concentrations, unless higher accuracy is needed. In order to have reliable results, it is however important to do the model and feature selection carefully to avoid overfitting and to use methods to detect the concept drift.

Efficient explorative data analysis systems must take into account both what a user knows and wants to know. This paper proposes a principled framework for interactive visual exploration of relations in data, through views most informative given the user's current knowledge and objectives. The user can input pre-existing knowledge of relations in the data and also formulate specific exploration interests, which are then taken into account in the exploration. The idea is to steer the exploration process towards the interests of the user, instead of showing uninteresting or already known relations. The user's knowledge is modelled by a distribution over data sets parametrised by subsets of rows and columns of data, called tile constraints. We provide a computationally efficient implementation of this concept based on constrained randomisation. Furthermore, we describe a novel dimensionality reduction method for finding the views most informative to the user, which at the limit of no background knowledge and with generic objectives reduces to PCA. We show that the method is suitable for interactive use and is robust to noise, outperforms standard projection pursuit visualisation methods, and gives understandable and useful results in analysis of real-world data. We provide an open-source implementation of the framework.

Regression analysis is a standard supervised machine learning method used to model an outcome variable in terms of a set of predictor variables. In most real-world applications the true value of the outcome variable we want to predict is unknown outside the training data, i.e., the ground truth is unknown. Phenomena such as overfitting and concept drift make it difficult to directly observe when the estimate from a model potentially is wrong. In this paper we present an efficient framework for estimating the generalization error of regression functions, applicable to any family of regression functions when the ground truth is unknown. We present a theoretical derivation of the framework and empirically evaluate its strengths and limitations. We find that it performs robustly and is useful for detecting concept drift in datasets in several real-world domains.

Running large-eddy simulations (LES) can be burdensome and computationally too expensive from the application point-of-view for example to support urban planning. In this study, regression models are used to replicate modelled air pollutant concentrations from LES in urban boulevards. We study the performance of regression models and discuss how to detect situations where the models are applied outside their training domain and their outputs cannot be trusted. Regression models from 10 different model families are trained and a cross-validation methodology is used to evaluate their performance and to find the best set of features needed to reproduce the LES outputs. We also test the regression models on an independent testing dataset. Our results suggest that in general, log-linear regression gives the best and most robust performance on new independent data. It clearly outperforms the dummy model which would predict constant concentrations for all locations (mRMSE of 0.76 vs 1.78 of the dummy model). Furthermore, we demonstrate that it is possible to detect concept drift, i.e., situations where the model is applied outside its training domain and a new LES run may be necessary to obtain reliable results. Regression models can be used to replace LES simulations in estimating air pollutant concentrations, unless higher accuracy is needed. In order to have reliable results, it is however important to do the model and feature selection carefully to avoid over-fitting and to use methods to detect the concept drift.

Motivated by Gromov’s subgroup conjecture (GSC), a fundamental open conjecture in the area of geometric group theory, we tackle the problem of the existence of partic- ular types of subgroups—arising from so-called periodic apartments—for a specific set of hyperbolic groups with respect to which GSC is currently open. This problem is equiv- alent to determining whether specific types of graphs with a non-trivial combination of properties exist. The existence of periodic apartments allows for ruling the groups out as some of the remaining potential counterexamples to GSC. Our approach combines both automated reasoning techniques—in particular, Boolean satisfiability (SAT) solving—with problem-specific orderly generation. Compared to earlier attempts to tackle the problem through computational means, our approach scales noticeably better, and allows for both confirming results from a previous computational treatment for smaller parameter values as well as ruling out further groups out as potential counterexamples to GSC.