Weight-Sharing Neural Architecture Search: A Battle to Shrink the Optimization Gap

Neural architecture search (NAS) has attracted increasing attention. In recent years, individual search methods have been replaced by weight-sharing search methods for higher search efficiency, but the latter methods often suffer lower instability. This article provides a literature review on these methods and owes this issue to the optimization gap . From this perspective, we summarize existing approaches into several categories according to their efforts in bridging the gap, and we analyze both advantages and disadvantages of these methodologies. Finally, we share our opinions on the future directions of NAS and AutoML. Due to the expertise of the authors, this article mainly focuses on the application of NAS to computer vision problems.

Distribution of the European bladdernut Staphylea pinnata (Staphyleaceae) in Poland

This paper presents the current distribution range of the protected shrub European bladdernut Staphylea pinnata in Poland. The study was based on a literature review, multidisciplinary search methods and field visits. The paper presents 211 sites considered as natural (including 48 new) and 143 sites considered as sites of anthropogenic origin (including 116 new). The data included in this paper extend the natural range of the species mainly in the Sudety Mts & Sudety Foreland, Rożnów Foothills, and the Tarnogród Plateau. It has been confirmed that the areas of Przemyśl Foothills (54 sites) and Dynów Foothills (49 sites) are the sites most abundant in bladdernut considered as natural. A few sites of anthropogenic origin have been found in the area of northern Poland, several hundred kilometres from the border of the bladdernut continual distribution range. In the case of sites of anthropogenic origin, knowledge of their distribution has increased significantly. This paper provides a basis for further research on the bladdernut in Poland and Europe and may be useful in population management of this protected species.

Breadth First Search and Inference Methods Integration to satisfy table constraints

Within the Constraint Programming technology, so-called table constraints such as typical tables, compressed tables, smart tables, segmented tables, etc, are widely used. They can be used to represent any other types of constraints, and algorithms of the table constraint propagation (logical inference on constraints) allow eliminating a lot of "redundant" values from the domains of variables, while having low computational complexity. In the previous studies, the author proposed to divide smart tables into structures of C- and D-types. The generally accepted methodology for solving con-straint satisfaction problems is the combined application of constraint propagation methods and backtracking depth-first search methods. In the study, it is proposed to integrate breadth-first search methods and author`s method of table con-straint propagation. D-type smart tables are proposed to be represented as a join of several orthogonalized C-type smart tables. The search step is to select a pair of C-type smart tables to be joined and then propagate the restrictions. To de-termine the order of joining orthogonalized smart tables at each step of the search, a specialized heuristic is used, which reduces the search space, taking into account further calculations. When the restrictions are extended, the acceleration of the computation process is achieved by applying the developed reduction rules for the case of C-type smart tables. The developed hybrid method allows one to find all solutions to the problems of satisfying constraints modeled using one or several D-type smart tables, without decomposing tabular constraints into elementary tuples.

Search Methods for Continuous Gravitational-Wave Signals from Unknown Sources in the Advanced-Detector Era

Continuous gravitational waves are long-lasting forms of gravitational radiation produced by persistent quadrupolar variations of matter. Standard expected sources for ground-based interferometric detectors are neutron stars presenting non-axisymmetries such as crustal deformations, r-modes or free precession. More exotic sources could include decaying ultralight boson clouds around spinning black holes. A rich suite of data-analysis methods spanning a wide bracket of thresholds between sensitivity and computational efficiency has been developed during the last decades to search for these signals. In this work, we review the current state of searches for continuous gravitational waves using ground-based interferometer data, focusing on searches for unknown sources. These searches typically consist of a main stage followed by several post-processing steps to rule out outliers produced by detector noise. So far, no continuous gravitational wave signal has been confidently detected, although tighter upper limits are placed as detectors and search methods are further developed.

Zeroth Order Policy Search Methods for Global Optimization Problems: An Experimental Study

Os métodos Policy Search (PS) vem sendo utilizados nos últimos anos para se aprender, automaticamente, algoritmos de otimização, obtendo resultados animadores. Neste trabalho, consideramos métodos PS para aprender algoritmos de otimização para problemas de otimização global, considerando um cenário pouco estudado: funções de alta dimensionalidade e os algoritmos de otimização não possuem acesso às derivadas da função a ser otimizada. Os resultados apontam, que apesar das dificuldades, os algoritmos de otimização aprendidos têm um desempenho promissor no cenário estudado.

One-Machine Scheduling with Time-Dependent Capacity via Efficient Memetic Algorithms

This paper addresses the problem of scheduling a set of jobs on a machine with time-varying capacity, with the goal of minimizing the total tardiness objective function. This problem arose in the context scheduling the charging times of a fleet of electric vehicles and it is NP-hard. Recent work proposed an efficient memetic algorithm for solving the problem, combining a genetic algorithm and a local search method. The local search procedure is based on swapping consecutive jobs on a C-path, defined as a sequence of consecutive jobs in a schedule. Building on it, this paper develops new memetic algorithms that stem from new local search procedures also proposed in this paper. The local search methods integrate several mechanisms to make them more effective, including a new condition for swapping pairs of jobs, a hill climbing approach, a procedure that operates on several C-paths and a method that interchanges jobs between different C-paths. As a result, the new local search methods enable the memetic algorithms to reach higher-quality solutions. Experimental results show significant improvements over existing approaches.