Browsing by Author "Barsbey, Melih."

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Perceptual simulations beyond the sentence: the effect of narrative shifts on perceptual simulations
    (Thesis (M.A.) - Bogazici University. Institute for Graduate Studies in Social Sciences, 2014., 2014.) Barsbey, Melih.; Sarıbay, S. Adil.
    The present research investigates perceptual simulations within the context of narratives, as only a minority of previous research that investigated mental simulations in language comprehension did so on the discursive level. Participants were given short narrative excerpts, in which random pictorial probes were placed asking participants to decide whether they had encountered the object depicted in the picture within the text. In critical trials, the object depicted was indeed present in the narratives, but the shape in which the object was depicted either matched or mismatched the shape implied by the text. It has previously been documented that people are slower and less accurate in responding to pictorial probes in the case of shape mismatch compared to shape match, implying that they had created a contextspecific perceptual simulation of the critical object as they encountered it in the text. This study extends previous research by investigating whether the match effect described above weakens when a spatial or temporal shift is introduced between the appearance of the object in the text and the pictorial probe. Besides the replication of the shape match effect in a narrative context, the results showed that there was mild evidence towards perceptual simulations getting weaker after a spatial shift. Interestingly, the reverse was the case for the temporal shift. Overall, the results imply that the investigation of mental simulations within the narrative context is a fruitful direction for further research in arriving at a better understanding of embodied language comprehension and language comprehension in general.
  • Thumbnail Image
    Item
    Utilizing nonnegative tensor factorization methods for inference, model selection, and analysis in supervised learning
    (Thesis (Ph.D.) - Bogazici University. Institute for Graduate Studies in Science and Engineering, 2023., 2023) Barsbey, Melih.; Özgür, Arzucan.; Cemgil, Ali Taylan.
    This thesis focuses on utilizing nonnegative tensor factorization (NTF) methods in various areas of supervised learning. We start with the introduction of a probabilistic NTF framework that can accommodate a wide range of modeling assumptions while maintaining algorithmic efficiency during inference. The flexibility provided by this framework is then utilized for inference, model selection, and analysis in various supervised learning problems. In the first of these scenarios, we use this approach to effectively model time series with nested, complex seasonalities, ensuring accuracy and interpretability. We then propose a novel method for learning to defer to an expert based on the output of a machine learning model in classification problems, and show that NTF can be utilized to extend this method to arbitrarily complex settings. Afterwards, we investigate when and why deep neural networks’ parameters become compressible, and use the aforementioned NTF framework to help analyze how these dynamics are reflected in the representation space. In addition to making independent contributions to various areas of supervised learning, our work shows that, coupled with a convenient modeling approach, NTF can be beneficial for a wide range of supervised learning problems. NOTE Keywords : Machine learning, Nonnegative tensor factorization, Graphical models, Deep learning.