Browsing by Author "Brandby, Johan"

Now showing 1 - 2 of 2

Att mäta SVT-program
(2019-07-01) Brandby, Johan; Wennerblom, Julia; Andersson, Stina; Gardell, Therese; University of Gothenburg/Department of Mathematical Science; Göteborgs universitet/Institutionen för matematiska vetenskaper
The purpose of this study is to examine whether or not the concept of informational advantage in series has a correlation with the popularity of the series. Informational advantage as a dramaturgical tool means that the viewer, at any given point throughout the series, either knows more, less or the same amount as the character in the show; these stages are called dramatic irony, mystery, and suspence respectively. Three different questions were examined. The time spent in one stage before switching to the next is examined with an Anderson-Darling test to see if it fits any statistical distribution, and if there is a difference between series with high and low viewership numbers. The conclusion is that it cannot be rejected that the time follows a log-normal distribution and that the expected time a program will spend in one stage before switching is shorter in programs with high viewership numbers. Furthermore, whether the total percentage of time spent in the different stages has a linear relationship with the viewership numbers is examined with linear regression using the method of least squares. With the collected data it was difficult to make a definitive conclusion, however the data implies that the relationship is stronger for the time spent in dramatic irony and viewership numbers, than the relationship for mystery. The relationship is especially stronger for the online viewership. Lastly it is examined whether the percentage of switches between the different states has a linear relationship with the viewership numbers, also with the method of least squares. The result points towards switches between dramatic irony and suspence correlating positively with viewership, while switches between mystery and suspence seem to correlate negatively with viewership.
An Empirical Survey of Bandits in an Industrial Recommender System Setting
(2023-09-21) Schwarz, Tobias; Brandby, Johan; Göteborgs universitet/Institutionen för data- och informationsteknik; University of Gothenburg/Department of Computer Science and Engineering
In this thesis, the effects of incorporating unstructured data—images in the wild—in contextual multi-armed bandits are investigated, when used within a recommender system setting, which focuses on picture-based content suggestion. The idea is to employ image features, extracted by a pre-trained convolutional neural network, and study the resulting bandit behaviors when including respective excluding this information in the typical context creation, which normally relies on structured data sources—such as metadata. The evaluation is made both online, through A/B-testing enabled by the industrial partner YouPic AB, and offline, effectuated by a simulation pipeline that models the online counterpart. The results are compiled as a survey, covering a selection of contextual bandit algorithms, highlighting the differences brought by the unstructured data. The offline result points towards that if the contextual bandit utilizes a joint or hybrid action-value function, with respect to the parameterization, the addition of the image vectors can significantly outperform the instances without it; however, if a disjoint model is instead employed, no noticeable change is observed. In comparison, those from the online trials can be interpreted as supporting the inclusion of convolutional features, but due to meager and unbalanced sample sizes, the outcomes are deemed inconclusive. To summarize, though there is support for incorporating unstructured data, given that the action-value function is joint or hybrid, the online experiments gave too little evidence for any trustworthy findings; in other words, the question is still partially open.