arXiv subjects

Mark Stalzer

Publications and source records attributed to Mark Stalzer.

5 recordsLinked to original sources

arXiv2018

Finding Exact Forms on Thermodynamic Manifolds

Because only two variables are needed to characterize a simple thermodynamic system in equilibrium, any such system is constrained on a 2D manifold. Of particular interest are the exact 1-forms on the cotangent space of that manifold, since the integral of exact 1-forms is path-independent, a crucial property satisfied by state variables such as the internal energy dE and the entropy dS. Our prior work shows that given an appropriate language of vector calculus, a machine can re-discover the Maxwell equations and the incompressible Navier-Stokes equations from simulation. We speculate that we can enhance this language by including differential forms. In this paper, we use the example of classical thermodynamics to show that there exists a simple algorithm to automate the process of finding exact 1-forms on a thermodynamic manifold. Since entropy appears in various fields of science in different guises, a potential extension of this work is to use the machinery developed in this paper to re-discover the expressions for entropy from data in fields other than classical thermodynamics.

math-ph↗

arXiv2017

Bayesian Optimization for Parameter Tuning of the XOR Neural Network

When applying Machine Learning techniques to problems, one must select model parameters to ensure that the system converges but also does not become stuck at the objective function's local minimum. Tuning these parameters becomes a non-trivial task for large models and it is not always apparent if the user has found the optimal parameters. We aim to automate the process of tuning a Neural Network, (where only a limited number of parameter search attempts are available) by implementing Bayesian Optimization. In particular, by assigning Gaussian Process Priors to the parameter space, we utilize Bayesian Optimization to tune an Artificial Neural Network used to learn the XOR function, with the result of achieving higher prediction accuracy.

stat.ML↗

arXiv2017

Deriving Compact Laws Based on Algebraic Formulation of a Data Set

In various subjects, there exist compact and consistent relationships between input and output parameters. Discovering the relationships, or namely compact laws, in a data set is of great interest in many fields, such as physics, chemistry, and finance. While data discovery has made great progress in practice thanks to the success of machine learning in recent years, the development of analytical approaches in finding the theory behind the data is relatively slow. In this paper, we develop an innovative approach in discovering compact laws from a data set. By proposing a novel algebraic equation formulation, we convert the problem of deriving meaning from data into formulating a linear algebra model and searching for relationships that fit the data. Rigorous proof is presented in validating the approach. The algebraic formulation allows the search of equation candidates in an explicit mathematical manner. Searching algorithms are also proposed for finding the governing equations with improved efficiency. For a certain type of compact theory, our approach assures convergence and the discovery is computationally efficient and mathematically precise.

cs.LG↗

arXiv2015

A Preliminary Review of Influential Works in Data-Driven Discovery

The Gordon and Betty Moore Foundation ran an Investigator Competition as part of its Data-Driven Discovery Initiative in 2014. We received about 1,100 applications and each applicant had the opportunity to list up to five influential works in the general field of "Big Data" for scientific discovery. We collected nearly 5,000 references and 53 works were cited at least six times. This paper contains our preliminary findings.

cs.DL↗

arXiv2014

Parallel implementation of fast randomized algorithms for the decomposition of low rank matrices

We analyze the parallel performance of randomized interpolative decomposition by decomposing low rank complex-valued Gaussian random matrices up to 64 GB. We chose a Cray XMT supercomputer as it provides an almost ideal PRAM model permitting quick investigation of parallel algorithms without obfuscation from hardware idiosyncrasies. We obtain that on non-square matrices performance becomes very good, with overall runtime over 70 times faster on 128 processors. We also verify that numerically discovered error bounds still hold on matrices nearly two orders of magnitude larger than those previously tested.

cs.DC↗