arXiv subjects

Stefan Wild

Publications and source records attributed to Stefan Wild.

8 recordsLinked to original sources

arXiv2020

Recursive Two-Step Lookahead Expected Payoff for Time-Dependent Bayesian Optimization

We propose a novel Bayesian method to solve the maximization of a time-dependent expensive-to-evaluate oracle. We are interested in the decision that maximizes the oracle at a finite time horizon, when relatively few noisy evaluations can be performed before the horizon. Our recursive, two-step lookahead expected payoff ($\texttt{r2LEY}$) acquisition function makes nonmyopic decisions at every stage by maximizing the estimated expected value of the oracle at the horizon. $\texttt{r2LEY}$ circumvents the evaluation of the expensive multistep (more than two steps) lookahead acquisition function by recursively optimizing a two-step lookahead acquisition function at every stage; unbiased estimators of this latter function and its gradient are utilized for efficient optimization. $\texttt{r2LEY}$ is shown to exhibit natural exploration properties far from the time horizon, enabling accurate emulation of the oracle, which is exploited in the final decision made at the horizon. To demonstrate the utility of $\texttt{r2LEY}$, we compare it with time-dependent extensions of popular myopic acquisition functions via both synthetic and real-world datasets.

math.OC↗

arXiv2019

Scalable Reinforcement-Learning-Based Neural Architecture Search for Cancer Deep Learning Research

Cancer is a complex disease, the understanding and treatment of which are being aided through increases in the volume of collected data and in the scale of deployed computing power. Consequently, there is a growing need for the development of data-driven and, in particular, deep learning methods for various tasks such as cancer diagnosis, detection, prognosis, and prediction. Despite recent successes, however, designing high-performing deep learning models for nonimage and nontext cancer data is a time-consuming, trial-and-error, manual task that requires both cancer domain and deep learning expertise. To that end, we develop a reinforcement-learning-based neural architecture search to automate deep-learning-based predictive model development for a class of representative cancer data. We develop custom building blocks that allow domain experts to incorporate the cancer-data-specific characteristics. We show that our approach discovers deep neural network architectures that have significantly fewer trainable parameters, shorter training time, and accuracy similar to or higher than those of manually designed architectures. We study and demonstrate the scalability of our approach on up to 1,024 Intel Knights Landing nodes of the Theta supercomputer at the Argonne Leadership Computing Facility.

cs.LG↗

arXiv2019

Few layer 2D pnictogens catalyze the alkylation of soft nucleophiles with esters

Group 15 elements in zero oxidation state (P, As, Sb and Bi), also called pnictogens, are rarely used in catalysis due to the difficulties associated in preparing well-structured and stable materials. Here, we report on the synthesis of highly exfoliated, few layer 2D phosphorene and antimonene in zero oxidation state, suspended in an ionic liquid, with the native atoms ready to interact with external reagents while avoiding aerobic or aqueous decomposition pathways, and on their use as efficient catalysts for the alkylation of nucleophiles with esters. The few layer pnictogen material circumvents the extremely harsh reaction conditions associated to previous superacid-catalyzed alkylations, by enabling an alternative mechanism on surface, protected from the water and air by the ionic liquid. These 2D catalysts allow the alkylation of a variety of acid-sensitive organic molecules and giving synthetic relevancy to the use of simple esters as alkylating agents.

physics.chem-ph↗

arXiv2019

Lattice Opening Upon Bulk Reductive Covalent Functionalization of Black Phosphorus

The chemical bulk reductive covalent functionalization of thin layer black phosphorus (BP) using BP intercalation compounds has been developed. Through effective reductive activation, covalent functionalization of the charged BP is achieved by organic alkyl halides. Functionalization was extensively demonstrated by means of several spectroscopic techniques and DFT calculations, showing higher functionalization degrees than the neutral routes.

physics.app-ph↗

arXiv2019

Monolayer black phosphorus by sequential wet-chemical surface oxidation

We report a straightforward chemical methodology for controlling the thickness of black phosphorus flakes down to the monolayer limit by layer-by-layer oxidation and thinning, using water as solubilizing agent. Moreover, the oxidation process can be stopped at will by two different passivation procedures, namely the non-covalent functionalization with perylene diimide chromophores, which prevents the photooxidation, or by using a protective ionic liquid layer. The obtained flakes preserve their electronic properties as demonstrated by fabricating a BP field-effect transistor (FET). This work paves the way for the preparation of BP devices with controlled thickness

physics.app-ph↗

arXiv2018

Effect of TCNQ layer cover on oxidation dynamics of black phosphorus

The puckered surface of black phosphorus represents an ideal substrate for an unconventional arrangement of physisorbed species and the resulting specific two-dimensional chemistry of this system. This opens the way to investigate the chemical and physical properties of locally confined areas of black phosphorus without the necessity for further physical downscaling of the material. We have evaporated TCNQ on top of black phosphorus under over-saturation non-equilibrium conditions in vacuum. The evolution of linear density and height of droplets formed through oxidation during exposure to air was studied time-dependently by scanning-force microscopy. Our study suggests that the TCNQ molecules spontaneously arrange in a thin layer of the order of a few nm height, which, however, is fragmented with a periodicity of about 100 nm. It is shown that within the confined space separating the layer fragments the chemical dynamics of the oxidation process is remarkably different than on a bare black phosphorus surface.

cond-mat.mtrl-sci↗

arXiv2018

Electronic and magnetic properties of black phosphorus

Black phosphorus has emerged as the next member in the graphene inspired two-dimensional materials family. Its electronic and magnetic properties are studied herein using electron and nuclear magnetic resonance techniques (ESR and NMR) and microwave conductivity measurement. The latter is a unique technique to study conductivity on air sensitive samples. The ESR study indicates the absence of free charge carriers and no sign of paramagnetic defects are found. $^{31}$P NMR shows the presence of a characteristic Pake doublet structure due to the interaction between $I = 1/2$ nuclei. Microwave conductivity shows, in accordance with the ESR results, that black phosphorus behaves as a semiconductor and we identify extrinsic and intrinsic charge carrier contributions to the conductivity and extracted the sizes of the gaps. ESR measurement also yields that bP might find applications as a microwave absorbent.

cond-mat.str-el↗

arXiv2015

Randomized Derivative-Free Optimization of Noisy Convex Functions

We propose STARS, a randomized derivative-free algorithm for unconstrained optimization when the function evaluations are contaminated with random noise. STARS takes dynamic, noise-adjusted smoothing step-sizes that minimize the least-squares error between the true directional derivative of a noisy function and its finite difference approximation. We provide a convergence rate analysis of STARS for solving convex problems with additive or multiplicative noise. Experimental results show that (1) STARS exhibits noise-invariant behavior with respect to different levels of stochastic noise; (2) the practical performance of STARS in terms of solution accuracy and convergence rate is significantly better than that indicated by the theoretical result; and (3) STARS outperforms a selection of randomized zero-order methods on both additive and multiplicative-noisy functions.

math.OC↗