05/09/2017 - 11:00am

05/09/2017 - 12:00pm

Speaker:

Gowri Srinivasan (LANL)

Abstract:

Microstructural information (fracture size, orientation, etc.) plays a key role in governing the dominant physics for two timely applications of interest to LANL: dynamic fracture processes like spall and fragmentation in metals (weapons performance) and detection of gas flow in static fractures in rock due to underground explosions (nuclear nonproliferation). Micro-fracture information is only known in a statistical sense, so representing millions of micro-fractures in 1000s of model runs to bound the uncertainty requires petabytes of information. Our critical advance is to integrate computational physics, machine learning and graph theory to make a paradigm shift from computationally intensive grid-based models to efficient graphs with at least 3 orders of magnitude speedup for Discrete Fracture Networks (DFNs).

Where:

CGU Math South

05/05/2017 - 1:00pm

05/05/2017 - 2:00pm

Speaker:

Kei Kobayashi (Fordham University)

Abstract:

Brownian motion has been employed to model a number of random time-dependent quantities observed in many different research areas. However, this classical model has several drawbacks; one notable shortcoming is that it does not allow the quantities to be constant over any time interval of positive length. One way to describe such constant periods is to introduce a random time change given by the so-called inverse stable subordinator. The Brownian motion composed with this specific time change is significantly different from the classical Brownian motion; for example, it is non-Markovian and has transition probability densities satisfying a time-fractional order heat equation.

Where:

CGU Math South

09/11/2017 - 4:15pm

09/11/2017 - 5:15pm

Speaker:

Brian Stock

Abstract:

TBA

Where:

TBA

05/01/2017 - 4:15pm

05/01/2017 - 5:15pm

Speaker:

Rebecca Garnett (NAWCWD China Lake)

Abstract:

With the advent of large-scale digital music repositories and personalized streaming radio software, there is a growing need for effective, autonomous methods of music categorization. The majority of published research in this area employ the physics of sound propagation and attempt to draw algorithmic parallels to the human auditory system for classification of music into different genres. However, deep neural network architectures are currently the state of the art for many classification problems. These deep networks typically require large amounts of data, long time scales, and extensive computational resources for training, putting constraints on their ability to be effectively implemented. Motivated by Mallat’s Invariant Scattering Convolution Networks (Bruna, Joan, and Stéphane Mallat. "Invariant scattering convolution networks." IEEE transactions on pattern analysis and machine intelligence 35.8 (2013): 1872-1886.), this work presents some preliminary studies to overcome these limitations. Mallat’s work demonstrated that respecting natural symmetries and adding robustness to deformations using non-linear functions can substantially improve classification. This study exploits these ideas to classify musical audio signals based on learned representations of their spectrograms’ dynamics. First, Nonnegative Matrix Factorization (NMF) was used to obtain a representation of spectrograms. Then the nonlinear max-pooling operator was used to add stability and reduce computational complexity. Finally, Hidden Markov Models (HMMs) were built to characterize the signal dynamics for each genre of music, and samples were classified according to how well they fit each HMM. Employing these HMMs induced a time-independent model, while the non-linear pooling step added robustness to deformations. Testing was executed against the well-studied GTZAN genre dataset and classification was performed using a multi-class Support Vector Machine (SVM). An 86% correct classification rate was achieved.

Where:

Emmy Noether Rm
Millikan 1021
Pomona College

03/06/2017 - 4:15pm

03/06/2017 - 5:10pm

Speaker:

Huiyi Hu (Google)

Abstract:

Deep learning uses artificial neural networks to uncover intricate pattern and structure in large data set. This type of method has achieved significant improvement in many fields such as visual object detection, object recognition, speech recognition and natural language processing (NLP) problems. As a result, it has been receiving rapidly increasing amount of attention from both academia and industry.

In this talk, I will first give an overview of the recent progress in deep learning along with a few examples. Then I will go into more technical details on what a basic deep learning model is made of and how it works (multi-layer neural networks and back propagation). At last I will present 1-2 examples in finer details, to show how deep learning technique is used in various applications.

Where:

Emmy Noether Rm
Millikan 1021
Pomona College

04/03/2017 - 4:15pm

04/03/2017 - 5:15pm

Speaker:

Kimberly Ayers (Pomona)

Abstract:

In this talk, we consider a finite family of dynamical systems all on the same compact metric space, M, and study what happens when switch between these systems at regular time intervals. We begin by isolating and examining the “switching” dynamics by constructing a space made up of piecewise constant functions, and then study the dynamics of this space under the left shift map. We demonstrate that this function space, when paired with the behavior on M, gives a skew product flow. We then define and generalize various recurrence and limit concepts for this new skew product flow that demonstrates hybrid continuous and discrete behavior.

Where:

Emmy Noether Rm
Millikan 1021
Pomona College

01/23/2017 - 4:15pm

01/23/2017 - 5:15pm

Speaker:

TBA

Abstract:

TBA

Where:

Emmy Noether Rm
Millikan 1021
Pomona College

04/24/2017 - 4:15pm

04/24/2017 - 5:15pm

Speaker:

Alan Krinik (Cal Poly Pomona)

Abstract:

We consider various recurrent birth-death chains on state space S1 = {0,1,2,...,H} and its associated dual birth-death chain on state space S2 = {−1,0,1,2,...,H} having absorbing states −1 and H. Assume P and P∗ are the one-step transition probability matrices of each birth-death chain respectively.

Conclusions:

1. P and P∗ have the same set of eigenvalues.

2. An explicit, simple formula for the eigenvalues of P (and P∗) is described as a function of H.

3. Pn and (P∗)n can be expressed exactly for n ∈ N.

Conclusion 2 follows from some nice linear algebra results on certain types of tridiagonal matrices found in Kouachi (2006, 2008). Our conclusion 3 has implications for ﬁnite-time gambler’s ruin problems. Some of the preceding results extend beyond birth-death chains to certain Markov chains. Our results for stochastic matrices suggest further ways to generalize Kouachi’s work. Explicit formula for ﬁnding transition probability functions of certain birth-death processes is also described.

Where:

Emmy Noether Rm
Millikan 1021
Pomona College

04/10/2017 - 4:15pm

04/10/2017 - 5:15pm

Speaker:

Chuntian Sharon Wang (UCLA)

Abstract:

Zakharov-Kuznetsov (ZK) equation is the long-wave small-amplitude limit of the Euler-Poisson system for cold plasma uniformly magnetized along one space direction. It is also a multi-dimensional extension of the Korteweg-de Vries (KdV) equation and a special case of the partially hyperbolic equations. The talk will focus on the well-posedness and regularity of both the deterministic and

Stochastic ZK equation, subjected to a rectangular domain in space dimensions two and three. Particularly, in the deterministic case, we obtain the global existence of strong solutions in 3D, which, for similar equations in fluid dynamics, is still open. For the stochastic ZK equation driven by a white noise, in 3D the existence of martingale solutions, and in 2D the uniqueness and existence of the pathwise solution are established, an analogy to the results of the weak solutions (in the PDE sense) in the deterministic case.

In terms of methodology, the focus is on the handling of the mixed features consisting of the partial hyperbolicity, nonlinearity, anisotropicity and stochasticity of the system, which, sitting at the interface among probability and analysis of the parabolic and hyperbolic PDEs, provides interesting and challenging mathematical complications.

Where:

Emmy Noether Rm
Millikan 1021
Pomona College

03/27/2017 - 4:15pm

03/27/2017 - 5:15pm

Speaker:

David Arnold (UCLA)

Abstract:

Spiral particle separators are devices used in the mining and

mineral processing industries to separate ores and clean coal. Their

design process remains fairly experimental, and the work I will speak

about aims to help improve mathematical modelling of the flow in these

devices, and other flow in helical geometries. The cross-sectional flow

is known to be an important factor for particle separation

characteristics, but measuring this is very difficult experimentally,

due to the small fluid depths associated with spiral separators. Using a

non-orthogonal coordinate system we derive the Navier-Stokes equations

and take the thin-film limit to obtain a simplified system of equations.

In this talk I will discuss the clear-fluid problem (with no particles

in the flow), for channels with arbitrary centreline pitch and radius,

and arbitrary (but shallow) cross-sectional shape. Remarkably, for

channels with rectangular cross-section, we are able to solve the

governing equations analytically. Finally, I will discuss modelling of

particle-laden flows in helical channels.

Where:

Emmy Noether Rm
Millikan 1021
Pomona College