The paper describes a simple, generic and yet highly accurate efficient importance sampling (EIS) Monte Carlo (MC) procedure for the evaluation of high-dimensional numerical integrals. EIS is based upon a sequence of auxiliary weighted regressions which actually are linear under appropriate conditions. It can be used to evaluate likelihood functions and byproducts thereof, such as ML estimators, for models which depend upon unobservable variables. A dynamic stochastic volatility model and a logit panel data model with unobserved heterogeneity (random effects) in both dimensions are used to provide illustrations of EIS high numerical accuracy, even under small number of MC draws. MC simulations are used to characterize the finite sample numerical and statistical properties of EIS-based ML estimators. r 2007 Elsevier B.V. All rights reserved.JEL classification: C15

The rapid development of large pre-trained language models has greatly increased the demand for model compression techniques, among which quantization is a popular solution. In this paper, we propose Binary-BERT, which pushes BERT quantization to the limit by weight binarization. We find that a binary BERT is hard to be trained directly than a ternary counterpart due to its complex and irregular loss landscape. Therefore, we propose ternary weight splitting, which initializes BinaryBERT by equivalently splitting from a half-sized ternary network. The binary model thus inherits the good performance of the ternary one, and can be further enhanced by fine-tuning the new architecture after splitting. Empirical results show that our Binary-BERT has only a slight performance drop compared with the full-precision model while being 24× smaller, achieving the state-of-the-art compression results on the GLUE and SQuAD benchmarks. (a) Full-precision Model. (b) Ternary Model. (c) Binary Model. (d) All Together.Figure 2: Loss landscapes visualization of the full-precision, ternary and binary models on MRPC. For (a), (b) and (c), we perturb the (latent) full-precision weights of the value layer in the 1 st and 2 nd Transformer layers, and compute their corresponding training loss. (d) shows the gap among the three surfaces by stacking them together. (a) MHA-QK. (b) MHA-V. (c) MHA-O. (d) FFN-Mid. (e) FFN-Out.

This work considers a pursuit-evasion game in which a number of pursuers are attempting to capture a single evader. Cooperation among multiple agents can be difficult to achieve, as it may require the selection of actions in the joint input space of all agents. This work presents a decentralized, real-time algorithm for cooperative pursuit of a single evader by multiple pursuers in bounded, simply-connected planar domains. The algorithm is based on minimizing the area of the generalized Voronoi partition of the evader. The pursuers share state information but compute their inputs independently. No assumptions are made about the evader's control strategies other than requiring the evader control inputs to conform to a speed limit. Proof of guaranteed capture is shown when the domain is convex and the players' motion models are kinematic. Simulation results are presented showing the efficiency and effectiveness of this strategy.

Deoxyribonucleic acid (DNA) is a biological macromolecule. Its main function is information storage. At present, the advancement of sequencing technology had caused DNA sequence data to grow at an explosive rate, which has also pushed the study of DNA sequences in the wave of big data. Moreover, machine learning is a powerful technique for analyzing largescale data and learns spontaneously to gain knowledge. It has been widely used in DNA sequence data analysis and obtained a lot of research achievements. Firstly, the review introduces the development process of sequencing technology, expounds on the concept of DNA sequence data structure and sequence similarity. Then we analyze the basic process of data mining, summary several major machine learning algorithms, and put forward the challenges faced by machine learning algorithms in the mining of biological sequence data and possible solutions in the future. Then we review four typical applications of machine learning in DNA sequence data: DNA sequence alignment, DNA sequence classification, DNA sequence clustering, and DNA pattern mining. We analyze their corresponding biological application background and significance, and systematically summarized the development and potential problems in the field of DNA sequence data mining in recent years. Finally, we summarize the content of the review and look into the future of some research directions for the next step.

Palladium diselenide (PdSe 2 ), a noble transition metal dichalcogenide has attracted increasing attention in recent years due to its outstanding semiconductor properties. In this study, 2D PdSe 2 nanofilms with thicknesses ranging from 2 to 28 nm and their heterostructures with Si substrates (PdSe 2 /Si heterostructures) were synthesized via a simple selenization method. Electrical transport characterizations based on field-effect transistor devices indicate that the few-layer PdSe 2 nanofilms exhibit a p-type semiconducting behavior. The optimal sensing performance of the PdSe 2 /Si-8 sensor (8 nm-thick PdSe 2 ) exhibits a comparable response toward NO 2 gas (ΔR/R a = ∼7.2% to 100 ppb and ∼18% to 1 ppm) at room temperature (RT). This response may result from the heterostructure effect and maze-type surface. Additionally, the PdSe 2 /Si-8 sensor has selectivity toward NO 2 compared with other gases including NO, H 2 , CO, NH 3 , and C 2 H 5 OH. Furthermore, density functional theory (DFT) calculations reveal the largest adsorption energy and charge transfer between NO 2 and the PdSe 2 surface, which coincides well with the experimental results. Moreover, the PdSe 2 /Si-8 sensor also exhibits repeatability and long-term stability during about 4 months at RT. These results indicate that the PdSe 2 /Si heterostructures may be a promising nanomaterial for room-temperature NO 2 gassensing devices.

No abstract

Abstract-Aggregation of a large number of responsive loads presents great power flexibility for demand response. An effective control and coordination scheme of flexible loads requires an accurate and tractable model that captures their aggregate flexibility. This paper proposes a novel approach to extract the aggregate flexibility of deferrable loads with heterogeneous parameters using polytopic projection approximation. First, an exact characterization of their aggregate flexibility is derived analytically, which in general contains exponentially many inequality constraints with respect to the number of loads. In order to have a tractable solution, we develop a numerical algorithm that gives a sufficient approximation of the exact aggregate flexibility. Geometrically, the flexibility of each individual load is a polytope, and their aggregation is the Minkowski sum of these polytopes. Our method originates from an alternative interpretation of the Minkowski sum as projection. The aggregate flexibility can be viewed as the projection of a high-dimensional polytope onto the subspace representing the aggregate power. We formulate a robust optimization problem to optimally approximate the polytopic projection with respect to the homothet of a given polytope. To enable efficient and parallel computation of the aggregate flexibility for a large number of loads, a muti-stage aggregation strategy is proposed. The scheduling policy for individual loads is also derived. Finally, an energy arbitrage problem is solved to demonstrate the effectiveness of the proposed method.

http://www.cs.tut.fi/~mirceanc/lysate_array_bioinformatics.htm