Source author record

James A. Yorke

James A. Yorke appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

Genomics math.DS nlin.CD Computer Vision Machine Learning math.CA math.FA Neural and Evolutionary Computing

Catalog footprint

What is connected

8works

8topics

4close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2022arXiv

Robustness of solutions of almost every system of equations

In mathematical modeling, it is common to have an equation $F(p)=c$ where the exact form of $F$ is not known. This article shows that there are large classes of $F$ where almost all $F$ share the same properties. The classes we investigate are vector spaces $\mathcal{F}$ of $C^1$ functions $F:\mathbb{R}^N \to \mathbb{R}^M$ that satisfy the following condition: $\mathcal{F}$ has ``almost constant rank'' (ACR) if there is a constant integer $ρ(\mathcal{F}) \geq 0$ such that rank$(DF(p))=ρ(\mathcal{F})$ for ``almost every'' $F\in \mathcal{F}$ and almost every $p\in\mathbb{R}^N$. If the vector space $\mathcal{F}$ is finite-dimensional, then ``almost every'' is with respect to Lebesgue measure on $\mathcal{F}$, and otherwise, it means almost every in the sense of prevalence, as described herein. Most function spaces commonly used for modeling purposes are ACR. In particular, we show that if all of the functions in $\mathcal{F}$ are linear or polynomial or real analytic, or if $\mathcal{F}$ is the set of all functions in a ``structured system'', then $\mathcal{F}$ is ACR. For each $F$ and $p$, the solution set of $p \in \mathbb{R}^N$ is SolSet$(p):= \{x: F(x)=F(p)\}.$ A solution set of $F(p)=c$ is called robust if it persists despite small changes in $F$ and $c$. The following two global results are proved for almost every $F$ in an ACR vector space $\mathcal{F}$: (1) Either the solution set SolSet$(p)$ is robust for almost every $p\in \mathbb{R}^N$, or none of the solution sets are robust. (2) The solution set SolSet$(p)$ is a $C^\infty$-manifold of dimension $d = N-ρ(\mathcal{F})$. In particular, $d$ is the same for almost every $F \in \mathcal{F}$.

preprint2020arXiv

Network Deconvolution

Convolution is a central operation in Convolutional Neural Networks (CNNs), which applies a kernel to overlapping regions shifted across the image. However, because of the strong correlations in real-world image data, convolutional kernels are in effect re-learning redundant data. In this work, we show that this redundancy has made neural network training challenging, and propose network deconvolution, a procedure which optimally removes pixel-wise and channel-wise correlations before the data is fed into each layer. Network deconvolution can be efficiently calculated at a fraction of the computational cost of a convolution layer. We also show that the deconvolution filters in the first layer of the network resemble the center-surround structure found in biological neurons in the visual regions of the brain. Filtering with such kernels results in a sparse representation, a desired property that has been missing in the training of neural networks. Learning from the sparse representation promotes faster convergence and superior results without the use of batch normalization. We apply our network deconvolution operation to 10 modern neural network models by replacing batch normalization within each. Extensive experiments show that the network deconvolution operation is able to deliver performance improvement in all cases on the CIFAR-10, CIFAR-100, MNIST, Fashion-MNIST, Cityscapes, and ImageNet datasets.

preprint2015arXiv

Quantitative Quasiperiodicity

The Birkhoff Ergodic Theorem concludes that time averages, that is, Birkhoff averages, $Σ_{n=1}^N f(x_n)/N$ of a function $f$ along an ergodic trajectory $(x_n)$ of a function $T$ converges to the space average $\int f dμ$, where $μ$ is the unique invariant probability measure. Convergence of the time average to the space average is slow. We introduce a modified average of $f(x_n)$ by giving very small weights to the "end" terms when $n$ is near $0$ or $N$. When $(x_n)$ is a trajectory on a quasiperiodic torus and $f$ and $T$ are $C^\infty$, we show that our weighted Birkhoff averages converge "super" fast to $\int f dμ$, {\em i.e.} with error smaller than every polynomial of $1/N$. Our goal is to show that our weighted Birkhoff average is a powerful computational tool, and this paper illustrates its use for several examples where the quasiperiodic set is one or two dimensional. In particular, we compute rotation numbers and conjugacies (i.e. changes of variables) and their Fourier series, often with 30-digit precision.

preprint2014arXiv

The Scaling of Chaos vs Periodicity: How Certain is it that an Attractor is Chaotic?

The character of the time-asymptotic evolution of physical systems can have complex, singular behavior with variation of a system parameter, particularly when chaos is involved. A perturbation of the parameter by a small amount $ε$ can convert an attractor from chaotic to non-chaotic or vice-versa. We call a parameter value where this can happen $ε$-uncertain. The probability that a random choice of the parameter is $ε$-uncertain commonly scales like a power law in $ε$. Surprisingly, two seemingly similar ways of defining this scaling, both of physical interest, yield different numerical values for the scaling exponent. We show why this happens and present a quantitative analysis of this phenomenon.

preprint2013arXiv

A hierarchical network heuristic for solving the orientation problem in genome assembly

In the past several years, the problem of genome assembly has received considerable attention from both biologists and computer scientists. An important component of current assembly methods is the scaffolding process. This process involves building ordered and oriented linear collections of contigs (continuous overlapping sequence reads) called scaffolds and relies on the use of mate pair data. A mate pair is a set of two reads that are sequenced from the ends of a single fragment of DNA, and therefore have opposite mutual orientations. When two reads of a mate-pair are placed into two different contigs, one can infer the mutual orientation of these contigs. While several orientation algorithms exist as part of assembly programs, all encounter challenges while solving the orientation problem due to errors from mis-assemblies in contigs or errors in read placements. In this paper we present an algorithm based on hierarchical clustering that independently solves the orientation problem and is robust to errors. We show that our algorithm can correctly solve the orientation problem for both faux (generated) assembly data and real assembly data for {\em R. sphaeroides bacteria}. We demonstrate that our algorithm is stable to both changes in the initial orientations as well as noise in the data, making it advantageous compared to traditional approaches.

preprint2013arXiv

QuorUM: an error corrector for Illumina reads

Motivation: Illumina Sequencing data can provide high coverage of a genome by relatively short (100 bp150 bp) reads at a low cost. Our goal is to produce trimmed and error-corrected reads to improve genome assemblies. Our error correction procedure aims at producing a set of error-corrected reads (1) minimizing the number of distinct false k-mers, i.e. that are not present in the genome, in the set of reads and (2) maximizing the number that are true, i.e. that are present in the genome. Because coverage of a genome by Illumina reads varies greatly from point to point, we cannot simply eliminate k-mers that occur rarely. Results: Our software, called QuorUM, provides reasonably accurate correction and is suitable for large data sets (1 billion bases checked and corrected per day per core). Availability: QuorUM is distributed as an independent software package and as a module of the MaSuRCA assembly software. Both are available under the GPL open source license at http://www.genome.umd.edu. Contact: gmarcais@umd.edu

preprint2010arXiv

Connecting period-doubling cascades to chaos

The appearance of infinitely-many period-doubling cascades is one of the most prominent features observed in the study of maps depending on a parameter. They are associated with chaotic behavior, since bifurcation diagrams of a map with a parameter often reveal a complicated intermingling of period-doubling cascades and chaos. Period doubling can be studied at three levels of complexity. The first is an individual period-doubling bifurcation. The second is an infinite collection of period doublings that are connected together by periodic orbits in a pattern called a cascade. It was first described by Myrberg and later in more detail by Feigenbaum. The third involves infinitely many cascades and a parameter value $μ_2$ of the map at which there is chaos. We show that often virtually all (i.e., all but finitely many) ``regular'' periodic orbits at $μ_2$ are each connected to exactly one cascade by a path of regular periodic orbits; and virtually all cascades are either paired -- connected to exactly one other cascade, or solitary -- connected to exactly one regular periodic orbit at $μ_2$. The solitary cascades are robust to large perturbations. Hence the investigation of infinitely many cascades is essentially reduced to studying the regular periodic orbits of $F(μ_2, \cdot)$. Examples discussed include the forced-damped pendulum and the double-well Duffing equation.

preprint1995arXiv

Bizarre topology is natural in dynamical systems

We describe an example of a $C^\infty$ diffeomorphism on a 7--manifold which has a compact invariant set such that uncountably many of its connected components are pseudocircles. (Any 7--manifold will suffice.) Furthermore, any diffeomorphism which is sufficiently close (in the $C^1$ metric) to the constructed map has a similar invariant set, and the dynamics of the map on the invariant set are chaotic.