In a previous mini-series of blog posts (see here, here, here, and here), I have introduced a small data set of 60 line drawings complemented with pairwise shape similarity ratings and analyzed this data set in the form of conceptual similarity spaces. Today, I will start a new mini-series about learning a mapping from images into these similarity spaces, following up on my prior work on the NOUN dataset (see here and here).
It’s about time for another blog post in my little “What is …?” series. Today I want to talk about a specific type of artificial neural network, namely convolutional neural networks (CNNs). CNNs are the predominant approach for classifying images and have already been implicitly used in my study on the NOUN data set as well as in the analysis of the Shape similarity ratings. With this blog post, I want to clarify the basic underlying structure of this type of neural networks.
I’m currently in the process of writing the background chapter on Machine Learning for my dissertation. In the context of doing that, I took a closer look at a widely used feature extraction technique called “Principle Component Analysis” (PCA). It can be described either on the intuitive level (“it finds orthogonal directions of greatest variance in the data”) or on the mathematical level (“it computes an eigenvalue decomposition of the data matrix”). What I found most difficult to understand was how these two descriptions are related to each other. This is essentially what I want to share in today’s blog post. Continue reading “What is a “Principle Component Analysis”?”
In the past, we have already talked about some machine learning models, including LTNs and β-VAE. Today, I would like to introduce the basic idea of linear support vector machines (SVMs) and how they can be useful for analyzing a conceptual space. Continue reading “What is a “Support Vector Machine”?”
This blog post closes the “A Hybrid Way: Reloaded” mini-series. So far, I have analyzed the MDS solutions in part 1 and investigated first regression results in part 2 (with respect to the effects of feature space, correct vs. shuffled targets, and regularization). Today, I want to analyze what happens if we use different MDS algorithms for constructing the similarity spaces and to what extent our regression results depend on the number of dimensions in the similarity space.