-
Differential Similarity in Higher Dimensional Spaces: Theory and Applications
Authors:
L. Thorne McCarty
Abstract:
This paper presents an extension and an elaboration of the theory of differential similarity, which was originally proposed in arXiv:1401.2411 [cs.LG]. The goal is to develop an algorithm for clustering and coding that combines a geometric model with a probabilistic model in a principled way. For simplicity, the geometric model in the earlier paper was restricted to the three-dimensional case. The…
▽ More
This paper presents an extension and an elaboration of the theory of differential similarity, which was originally proposed in arXiv:1401.2411 [cs.LG]. The goal is to develop an algorithm for clustering and coding that combines a geometric model with a probabilistic model in a principled way. For simplicity, the geometric model in the earlier paper was restricted to the three-dimensional case. The present paper removes this restriction, and considers the full $n$-dimensional case. Although the mathematical model is the same, the strategies for computing solutions in the $n$-dimensional case are different, and one of the main purposes of this paper is to develop and analyze these strategies. Another main purpose is to devise techniques for estimating the parameters of the model from sample data, again in $n$ dimensions. We evaluate the solution strategies and the estimation techniques by applying them to two familiar real-world examples: the classical MNIST dataset and the CIFAR-10 dataset.
△ Less
Submitted 10 May, 2024; v1 submitted 10 February, 2019;
originally announced February 2019.
-
Detection of a 6.7 GHz methanol kilomaser toward NGC4945
Authors:
Simon Ellingsen,
Tiege McCarty,
Shari Breen,
Maxim Voronkov
Abstract:
We report the detection of emission from the 6.7 GHz 5(1)-6(0)A+ transition of methanol towards the center of the nearby galaxy NGC4945. This is the first detection of emission in this transition beyond the local group. The isotropic luminosity of the integrated 6.7 GHz methanol emission is approximately a factor of 10000 greater than that for 6.7 GHz methanol masers associated with Galactic high-…
▽ More
We report the detection of emission from the 6.7 GHz 5(1)-6(0)A+ transition of methanol towards the center of the nearby galaxy NGC4945. This is the first detection of emission in this transition beyond the local group. The isotropic luminosity of the integrated 6.7 GHz methanol emission is approximately a factor of 10000 greater than that for 6.7 GHz methanol masers associated with Galactic high-mass star formation regions. The methanol emission is resolved on scales smaller than 40 pc and it appears unlikely that it could be due to a large concentration of Galactic-style star formation masers within a small region. Comparison with observations of other methanol transitions suggests that the 6.7 GHz methanol emission is due to a diffuse, low-gain maser, amplifying the background continuum radiation from the nuclear region. The methanol emission is blueshifted with respect to the the systemic velocity of the galaxy by several hundred kilometers per second and lies outside the velocity range associated with the dense gas and neutral hydrogen in the central region of NGC4945. We speculate that it may be associated with gas entrained in a superwind outflow from the nuclear region.
△ Less
Submitted 23 October, 2018; v1 submitted 22 October, 2018;
originally announced October 2018.
-
Clustering, Coding, and the Concept of Similarity
Authors:
L. Thorne McCarty
Abstract:
This paper develops a theory of clustering and coding which combines a geometric model with a probabilistic model in a principled way. The geometric model is a Riemannian manifold with a Riemannian metric, ${g}_{ij}({\bf x})$, which we interpret as a measure of dissimilarity. The probabilistic model consists of a stochastic process with an invariant probability measure which matches the density of…
▽ More
This paper develops a theory of clustering and coding which combines a geometric model with a probabilistic model in a principled way. The geometric model is a Riemannian manifold with a Riemannian metric, ${g}_{ij}({\bf x})$, which we interpret as a measure of dissimilarity. The probabilistic model consists of a stochastic process with an invariant probability measure which matches the density of the sample input data. The link between the two models is a potential function, $U({\bf x})$, and its gradient, $\nabla U({\bf x})$. We use the gradient to define the dissimilarity metric, which guarantees that our measure of dissimilarity will depend on the probability measure. Finally, we use the dissimilarity metric to define a coordinate system on the embedded Riemannian manifold, which gives us a low-dimensional encoding of our original data.
△ Less
Submitted 16 May, 2018; v1 submitted 10 January, 2014;
originally announced January 2014.
