-
Unlocking the potential of deep learning for marine ecology: overview, applications, and outlook
Authors:
Morten Goodwin,
Kim Tallaksen Halvorsen,
Lei Jiao,
Kristian Muri Knausgård,
Angela Helen Martin,
Marta Moyano,
Rebekah A. Oomen,
Jeppe Have Rasmussen,
Tonje Knutsen Sørdalen,
Susanna Huneide Thorbjørnsen
Abstract:
The deep learning revolution is touching all scientific disciplines and corners of our lives as a means of harnessing the power of big data. Marine ecology is no exception. These new methods provide analysis of data from sensors, cameras, and acoustic recorders, even in real time, in ways that are reproducible and rapid. Off-the-shelf algorithms can find, count, and classify species from digital i…
▽ More
The deep learning revolution is touching all scientific disciplines and corners of our lives as a means of harnessing the power of big data. Marine ecology is no exception. These new methods provide analysis of data from sensors, cameras, and acoustic recorders, even in real time, in ways that are reproducible and rapid. Off-the-shelf algorithms can find, count, and classify species from digital images or video and detect cryptic patterns in noisy data. Using these opportunities requires collaboration across ecological and data science disciplines, which can be challenging to initiate. To facilitate these collaborations and promote the use of deep learning towards ecosystem-based management of the sea, this paper aims to bridge the gap between marine ecologists and computer scientists. We provide insight into popular deep learning approaches for ecological data analysis in plain language, focusing on the techniques of supervised learning with deep neural networks, and illustrate challenges and opportunities through established and emerging applications of deep learning to marine ecology. We use established and future-looking case studies on plankton, fishes, marine mammals, pollution, and nutrient cycling that involve object detection, classification, tracking, and segmentation of visualized data. We conclude with a broad outlook of the field's opportunities and challenges, including potential technological advances and issues with managing complex data sets.
△ Less
Submitted 29 September, 2021;
originally announced September 2021.
-
Completely Random Measures and Lévy Bases in Free probability
Authors:
Francesca Collet,
Fabrizio Leisen,
Steen Thorbjørnsen
Abstract:
This paper develops a theory for completely random measures in the framework of free probability. A general existence result for free completely random measures is established, and in analogy to the classical work of Kingman it is proved that such random measures can be decomposed into the sum of a purely atomic part and a (freely) infinitely divisible part. The latter part (termed a free Lévy bas…
▽ More
This paper develops a theory for completely random measures in the framework of free probability. A general existence result for free completely random measures is established, and in analogy to the classical work of Kingman it is proved that such random measures can be decomposed into the sum of a purely atomic part and a (freely) infinitely divisible part. The latter part (termed a free Lévy basis) is studied in detail in terms of the free Lévy-Khintchine representation and a theory parallel to the classical work of Rajput and Rosinski is developed. Finally a Lévy-Itô type decomposition for general free Lévy bases is established.
△ Less
Submitted 10 July, 2020;
originally announced July 2020.
-
Uffe Haagerup - His life and mathematics
Authors:
Mohammad Sal Moslehian,
Erling Størmer,
Steen Thorbjørnsen,
Carl Winsløw
Abstract:
In remembrance of Professor Uffe Valentin Haagerup (1949--2015), as a brilliant mathematician, we review some aspects of his life, and his outstanding mathematical accomplishments.
In remembrance of Professor Uffe Valentin Haagerup (1949--2015), as a brilliant mathematician, we review some aspects of his life, and his outstanding mathematical accomplishments.
△ Less
Submitted 12 September, 2017;
originally announced September 2017.
-
The normal distribution is freely selfdecomposable
Authors:
Takahiro Hasebe,
Noriyoshi Sakuma,
Steen Thorbjørnsen
Abstract:
The class of selfdecomposable distributions in free probability theory was introduced by Barndorff-Nielsen and the third named author. It constitutes a fairly large subclass of the freely infinitely divisible distributions, but so far specific examples have been limited to Wigner's semicircle distributions, the free stable distributions, two kinds of free gamma distributions and a few other exampl…
▽ More
The class of selfdecomposable distributions in free probability theory was introduced by Barndorff-Nielsen and the third named author. It constitutes a fairly large subclass of the freely infinitely divisible distributions, but so far specific examples have been limited to Wigner's semicircle distributions, the free stable distributions, two kinds of free gamma distributions and a few other examples. In this paper, we prove that the (classical) normal distributions are freely selfdecomposable. More generally it is established that the Askey-Wimp-Kerov distribution $μ_c$ is freely selfdecomposable for any $c$ in $[-1,0]$. The main ingredient in the proof is a general characterization of the freely selfdecomposable distributions in terms of the derivative of their free cumulant transform.
△ Less
Submitted 20 July, 2017; v1 submitted 2 January, 2017;
originally announced January 2017.
-
Unimodality of the freely selfdecomposable probability laws
Authors:
Takahiro Hasebe,
Steen Thorbjornsen
Abstract:
We show that any freely selfdecomposable probability law is unimodal. This is the free probabilistic analog of Yamazato's result in [Ann. Probab. 6 (1978), 523-531].
We show that any freely selfdecomposable probability law is unimodal. This is the free probabilistic analog of Yamazato's result in [Ann. Probab. 6 (1978), 523-531].
△ Less
Submitted 6 January, 2015; v1 submitted 26 September, 2013;
originally announced September 2013.
-
On the free Gamma distributions
Authors:
Uffe Haagerup,
Steen Thorbjornsen
Abstract:
For each positive number $α$ we study the analog $ν_alpha$ in free probability of the classical Gamma distribution with parameter $α$. We prove that $ν_α$ is absolutely continuous and establish the main properties of the density, including analyticity and unimodality. We study further the asymptotic behavior of $ν_α$ as $α\downarrow0$.
For each positive number $α$ we study the analog $ν_alpha$ in free probability of the classical Gamma distribution with parameter $α$. We prove that $ν_α$ is absolutely continuous and establish the main properties of the density, including analyticity and unimodality. We study further the asymptotic behavior of $ν_α$ as $α\downarrow0$.
△ Less
Submitted 15 February, 2013;
originally announced February 2013.
-
Asymptotic expansions for the Gaussian Unitary Ensemble
Authors:
Uffe Haagerup,
Steen Thorbjørnsen
Abstract:
Let g:{\mathbb R} --> {\mathbb C} be a C^{\infty}-function with all derivatives bounded and let tr_n denote the normalized trace on the n x n matrices. In the paper [EM] Ercolani and McLaughlin established asymptotic expansions of the mean value E{tr_n(g(X_n))} for a rather general class of random matrices X_n,including the Gaussian Unitary Ensemble (GUE). Using an analytical approach, we provide…
▽ More
Let g:{\mathbb R} --> {\mathbb C} be a C^{\infty}-function with all derivatives bounded and let tr_n denote the normalized trace on the n x n matrices. In the paper [EM] Ercolani and McLaughlin established asymptotic expansions of the mean value E{tr_n(g(X_n))} for a rather general class of random matrices X_n,including the Gaussian Unitary Ensemble (GUE). Using an analytical approach, we provide in the present paper an alternative proof of this asymptotic expansion in the GUE case. Specifically we derive for a GUE random matrix X_n that
E{tr_n(g(X_n))}= \frac{1}{2π}\int_{-2}^2 g(x)\sqrt{4-x^2} dx +\sum_{j=1}^k\frac{α_j(g)}{n^{2j}}+ O(n^{-2k-2}),
where k is an arbitrary positive integer. Considered as mappings of g, we determine the coefficients α_j(g), j\in{\mathbb N}, as distributions (in the sense of L. Schwarts). We derive a similar asymptotic expansion for the covariance Cov{Tr_n[f(X_n)],Tr_n[g(X_n)]}, where f is a function of the same kind as g, and Tr_n=n tr_n. Special focus is drawn to the case where g(x)=1/(z-x) and f(x)=1/(w-x) for non-real complex numbers z and w. In this case the mean and covariance considered above correspond to, respectively, the one- and two-dimensional Cauchy (or Stieltjes) transform of the GUE(n,1/n).
△ Less
Submitted 23 September, 2010; v1 submitted 20 April, 2010;
originally announced April 2010.
-
A Random Matrix Approach to the Lack of Projections in C*_red(F_2)
Authors:
Uffe Haagerup,
Hanne Schultz,
Steen Thorbjornsen
Abstract:
In 1982 Pimsner and Voiculescu computed the K_0- and K_1-groups of the reduced group C*-algebra C*_red(F_k) of the free group F_k on k generators and settled thereby a long standing conjecture: C*_red(F_k) has no projections except for the trivial projections 0 and 1. Later simpler proofs of this conjecture were found by methods from K-theory or from non-commutative differential geometry. In thi…
▽ More
In 1982 Pimsner and Voiculescu computed the K_0- and K_1-groups of the reduced group C*-algebra C*_red(F_k) of the free group F_k on k generators and settled thereby a long standing conjecture: C*_red(F_k) has no projections except for the trivial projections 0 and 1. Later simpler proofs of this conjecture were found by methods from K-theory or from non-commutative differential geometry. In this paper we provide a new proof of the fact that C*_red(F_k) is projectionless. The new proof is based on random matrices and is obtained by a refinement of the methods recently used by the first and the third named author to show that the semigroup Ext(C*_red(F_k)) is not a group for k >= 2. By the same type of methods we also obtain that two phenomena proved by Bai and Silverstein for certain classes of random matrices: ``no eigenvalues outside (a small neighbourhood of) the support of the limiting distribution'' and ``exact separation of eigenvalues by gaps in the limiting distribution'' also hold for arbitrary non-commutative selfadjoint polynomials of independent GUE, GOE or GSE random matrices with matrix coefficients.
△ Less
Submitted 30 December, 2004;
originally announced December 2004.
-
A new application of Random Matrices: Ext(C*_{red}(F_2)) is not a group
Authors:
Uffe Haagerup,
Steen Thorbjornsen
Abstract:
In the process of developing the theory of free probability and free entropy, Voiculescu introduced in 1991 a random matrix model for a free semicircular system. Since then, random matrices have played a key role in von Neumann algebra theory (cf. [V8], [V9]). The main result of this paper is the following extension of Voiculescu's random matrix result: Let X_1^(n),...,X_r^(n) be a system of r s…
▽ More
In the process of developing the theory of free probability and free entropy, Voiculescu introduced in 1991 a random matrix model for a free semicircular system. Since then, random matrices have played a key role in von Neumann algebra theory (cf. [V8], [V9]). The main result of this paper is the following extension of Voiculescu's random matrix result: Let X_1^(n),...,X_r^(n) be a system of r stochastically independent n by n Gaussian self-adjoint random matrices as in Voiculescu's random matrix paper [V4], and let (x_1,...,x_r) be a semi-circular system in a C*-probability space. Then for every polynomial p in r noncommuting variables lim_{n->oo}||p(X_1^(n),...,X_r^(n))|| = ||p(x_1,...,x_r)||, for almost all omega in the underlying probability space. We use the result to show that the Ext-invariant for the reduced C*-algebra of the free group on 2 generators is not a group but only a semi-group. This problem has been open since Anderson in 1978 found the first example of a C*-algebra A for which Ext(A) is not a group.
△ Less
Submitted 21 June, 2007; v1 submitted 19 December, 2002;
originally announced December 2002.
