-
It's easy to fool yourself: Case studies on identifying bias and confounding in bio-medical datasets
Authors:
Subhashini Venugopalan,
Arunachalam Narayanaswamy,
Samuel Yang,
Anton Geraschenko,
Scott Lipnick,
Nina Makhortova,
James Hawrot,
Christine Marques,
Joao Pereira,
Michael Brenner,
Lee Rubin,
Brian Wainger,
Marc Berndl
Abstract:
Confounding variables are a well known source of nuisance in biomedical studies. They present an even greater challenge when we combine them with black-box machine learning techniques that operate on raw data. This work presents two case studies. In one, we discovered biases arising from systematic errors in the data generation process. In the other, we found a spurious source of signal unrelated…
▽ More
Confounding variables are a well known source of nuisance in biomedical studies. They present an even greater challenge when we combine them with black-box machine learning techniques that operate on raw data. This work presents two case studies. In one, we discovered biases arising from systematic errors in the data generation process. In the other, we found a spurious source of signal unrelated to the prediction task at hand. In both cases, our prediction models performed well but under careful examination hidden confounders and biases were revealed. These are cautionary tales on the limits of using machine learning techniques on raw data from scientific experiments.
△ Less
Submitted 6 April, 2020; v1 submitted 12 December, 2019;
originally announced December 2019.
-
A "bottom up" characterization of smooth Deligne-Mumford stacks
Authors:
Anton Geraschenko,
Matthew Satriano
Abstract:
In casual discussion, a stack is often described as a variety (the coarse space) together with stabilizer groups attached to some of its subvarieties. However, this description does not uniquely specify the stack. Our main result shows that for a large class of stacks one typically encounters, this description does indeed characterize them. Moreover, we prove that each such stack can be described…
▽ More
In casual discussion, a stack is often described as a variety (the coarse space) together with stabilizer groups attached to some of its subvarieties. However, this description does not uniquely specify the stack. Our main result shows that for a large class of stacks one typically encounters, this description does indeed characterize them. Moreover, we prove that each such stack can be described in terms of two simple procedures applied iteratively to its coarse space: canonical stack constructions and root stack constructions.
More precisely, if $\mathcal X$ is a smooth separated tame Deligne-Mumford stack of finite type over a field $k$ with trivial generic stabilizer, it is completely determined by its coarse space $X$ and the ramification divisor (on $X$) of the coarse space morphism $π\colon \mathcal X \to X$. Therefore, to specify such a stack, it is enough to specify a variety and the orders of the stabilizers of codimension 1 points. The group structures, as well as the stabilizer groups of higher codimension points, are then determined.
△ Less
Submitted 18 March, 2015;
originally announced March 2015.
-
There is no degree map for 0-cycles on Artin stacks
Authors:
Dan Edidin,
Anton Geraschenko,
Matthew Satriano
Abstract:
We show that there is no way to define degrees of 0-cycles on Artin stacks with proper good moduli spaces so that (i) the degree of an ordinary point is non-zero, and (ii) degrees are compatible with closed immersions.
We show that there is no way to define degrees of 0-cycles on Artin stacks with proper good moduli spaces so that (i) the degree of an ordinary point is non-zero, and (ii) degrees are compatible with closed immersions.
△ Less
Submitted 15 August, 2012;
originally announced August 2012.
-
Formal GAGA for good moduli spaces
Authors:
Anton Geraschenko,
David Zureick-Brown
Abstract:
We prove formal GAGA for good moduli space morphisms under an assumption of "enough vector bundles" (which holds for instance for quotient stacks). This supports the philosophy that though they are non-separated, good moduli space morphisms largely behave like proper morphisms.
We prove formal GAGA for good moduli space morphisms under an assumption of "enough vector bundles" (which holds for instance for quotient stacks). This supports the philosophy that though they are non-separated, good moduli space morphisms largely behave like proper morphisms.
△ Less
Submitted 1 July, 2015; v1 submitted 14 August, 2012;
originally announced August 2012.
-
Torus Quotients as Global Quotients by Finite Groups
Authors:
Anton Geraschenko,
Matthew Satriano
Abstract:
This article is motivated by the following local-to-global question: is every variety with tame quotient singularities globally the quotient of a smooth variety by a finite group? We show that this question has a positive answer for all quasi-projective varieties which are expressible as a quotient of a smooth variety by a split torus (e.g. simplicial toric varieties). Although simplicial toric va…
▽ More
This article is motivated by the following local-to-global question: is every variety with tame quotient singularities globally the quotient of a smooth variety by a finite group? We show that this question has a positive answer for all quasi-projective varieties which are expressible as a quotient of a smooth variety by a split torus (e.g. simplicial toric varieties). Although simplicial toric varieties are rarely toric quotients of smooth varieties by finite groups, we give an explicit procedure for constructing the quotient structure using toric techniques.
This result follow from a characterization of varieties which are expressible as the quotient of a smooth variety by a split torus. As an additional application of this characterization, we show that a variety with abelian quotient singularities may fail to be a quotient of a smooth variety by a finite abelian group. Concretely, we show that $\mathbb{P}^2/A_5$ is not expressible as a quotient of a smooth variety by a finite abelian group.
△ Less
Submitted 13 November, 2015; v1 submitted 23 January, 2012;
originally announced January 2012.
-
Toric Stacks II: Intrinsic Characterization of Toric Stacks
Authors:
Anton Geraschenko,
Matthew Satriano
Abstract:
The purpose of this paper and its prequel (Toric Stacks I) is to introduce and develop a theory of toric stacks which encompasses and extends the notions of toric stacks defined in [Laf02, BCS05, FMN10, Iwa09, Sat12, Tyo12], as well as classical toric varieties.
While the focus of the prequel is on how to work with toric stacks, the focus of this paper is how to show a stack is toric. For toric…
▽ More
The purpose of this paper and its prequel (Toric Stacks I) is to introduce and develop a theory of toric stacks which encompasses and extends the notions of toric stacks defined in [Laf02, BCS05, FMN10, Iwa09, Sat12, Tyo12], as well as classical toric varieties.
While the focus of the prequel is on how to work with toric stacks, the focus of this paper is how to show a stack is toric. For toric varieties, a classical result says that any normal variety with an action of a dense open torus arises from a fan. In [FMN09, Theorem 7.24], it is shown that a smooth separated DM stack with an action of a dense open stacky torus arises from a stacky fan. In the same spirit, the main result of this paper is that any Artin stack with an action of a dense open torus arises from a stacky fan under reasonable hypotheses.
△ Less
Submitted 5 August, 2014; v1 submitted 10 July, 2011;
originally announced July 2011.
-
Toric Stacks I: The Theory of Stacky Fans
Authors:
Anton Geraschenko,
Matthew Satriano
Abstract:
The purpose of this paper and its sequel (Toric Stacks II) is to introduce and develop a theory of toric stacks which encompasses and extends the notions of toric stacks defined in [Laf02, BCS05, FMN10, Iwa09, Sat12, Tyo12], as well as classical toric varieties.
In this paper, we define a \emph{toric stack} as a quotient of a toric variety by a subgroup of its torus (we also define a generically…
▽ More
The purpose of this paper and its sequel (Toric Stacks II) is to introduce and develop a theory of toric stacks which encompasses and extends the notions of toric stacks defined in [Laf02, BCS05, FMN10, Iwa09, Sat12, Tyo12], as well as classical toric varieties.
In this paper, we define a \emph{toric stack} as a quotient of a toric variety by a subgroup of its torus (we also define a generically stacky version). Any toric stack arises from a combinatorial gadget called a \emph{stacky fan}. We develop a dictionary between the combinatorics of stacky fans and the geometry of toric stacks, stressing stacky phenomena such as canonical stacks and good moduli space morphisms.
We also show that smooth toric stacks carry a moduli interpretation extending the usual moduli interpretations of $\mathbb{P}^n$ and $[\mathbb{A}^1/\mathbb{G}_m]$. Indeed, smooth toric stacks precisely solve moduli problems specified by (generalized) effective Cartier divisors with given linear relations and given intersection relations. Smooth toric stacks therefore form a natural closure to the class of moduli problems introduced for smooth toric varieties and smooth toric DM stacks in [Cox95] and [Per08], respectively.
We include a plethora of examples to illustrate the general theory. We hope that this theory of toric stacks can serve as a companion to an introduction to stacks, in much the same way that toric varieties can serve as a companion to an introduction to schemes.
△ Less
Submitted 5 August, 2014; v1 submitted 10 July, 2011;
originally announced July 2011.
