When Does Communication Learning Need Hierarchical Multi-Agent Deep Reinforcement Learning

Abstract

Multi-agent systems need to communicate to coordinate a shared task. We show that a recurrent neural network (RNN) can learn a communication protocol for coordination, even if the actions to coordinate are performed steps after the communication phase. We show that a separation of tasks with different temporal scale is necessary for successful learning. We contribute a hierarchical deep reinforcement learning model for multi-agent systems that separates the communication and coordination task from the action picking through a hierarchical policy. We further on show, that a separation of concerns in communication is beneficial but not necessary. As a testbed, we propose the Dungeon Lever Game and we extend the Differentiable Inter-Agent Learning (DIAL) framework. We present and compare results from different model variations on the Dungeon Lever Game.     

Specification methods for material-handling control algorithms in flexible manufacturing systems

Good methods are needed to specify, test, and debug material-handling control logic. This article surveys a number of representative methods for defining and describing control algorithms for programmable material-handling equipment used in flexible manufacturing systems. The methods are evaluated with regard to their suitability for communication between people and as bases for interfaces to automatic program generators. It is concluded that no single method is entirely satisfactory. Three methods (position diagrams, function block diagrams, and operation networks) have potential to be combined into an effective hybrid approach that minimizes the need for the user to switch between various conceptual models.     

Has risk associated with smoking increased? Results from the Copenhagen Center for Prospective Population Studies

BACKGROUND: Two recent much cited publications have raised the concern that risk associated with cigarette smoking has so far been underestimated. In this study we wish to determine whether excess all-cause mortality associated with smoking has increased during the last 20-30 years in a study population representative of the general Danish population and whether any such changes relate to changes in smoking behaviour. METHODS: Pooled data from three prospective population studies conducted in Copenhagen with detailed information on smoking habits. A total of 31,194 subjects, 17,669 males and 13,525 females, initially examined between 1964 and 1992 with examinations repeated at intervals from 1-10 years, were followed until 1995 for all-cause mortality. Relative mortality risk in smokers versus never-smokers was calculated within periods of five calendar years and compared throughout the study period. RESULTS: Male smokers' exposure did not change during the study period whereas female smokers' exposure to tobacco increased in terms of age at smoking onset, quantity smoked and depth of inhalation. During follow-up 5744 males and 2900 females died. In males, death rate ratios (comparing continuous smokers with never-smokers) did not change in the study period. In females, ratios increased from 1964-1978 to 1979-1994 by a factor of 1.3 (95% confidence interval 1.0-1.8). CONCLUSIONS: In agreement with the observed changes in smoking habits, excess mortality in male smokers did not increase whereas excess mortality in female smokers increased slightly.     

A process oriented semantics of the PRAM-language FORK

The parallel language FORK [1], based on a scalable shared memory model, is a PASCAL-like language with some additional parallel constructs. A PRAM (Parallel Random Access Machine) algorithm can be expressed on a high level of abstraction as a FORK program which is translated into efficient PRAM code guaranteeing theoretically predicted runtimes.

In this paper, we concentrate on those features of the language FORK related to parallelism, such as the group concept, a shared memory access and synchronous or asynchronous execution. We present a trace-based denotational interleaving semantics where processes describe synchronous computations. Processes are created or deleted dynamically and run asynchronously. Interleaving rules reflect the underlying CRCW (concurrent-read-concurrent-write) PRAM model.     

Common region of ALL-1 gene disrupted in epipodophyllotoxin-related secondary acute myeloid leukemia

Translocations at chromosomal band 11q23 characterize most de novo acute lymphoblastic leukemias (ALL) of infants, acute myeloid leukemias (AML) of infants and young children, and secondary AMLs following epipodophyllotoxin exposure. The chromosomal breakpoints at 11q23 have been cloned from isolated cases of de novo ALL and AML. Using an 859-base pair BamHI fragment of human ALL-1 complementary DNA that recognizes the genomic breakpoint region for de novo ALL and AML, we investigated two cases of secondary AML that followed etoposide-treated primary B-lineage ALL. In the first case, the translocation occurred between chromosomes 9 and 11 and the breakpoint at 11q23 localized to the same 9-kilobase region of the ALL-1 gene that is disrupted in most of the de novo leukemias. In the second case the translocation was between chromosomes 11 and 19. The breakpoint occurred outside of the ALL-1 breakpoint cluster region.     

Uncertainty in Quantitative Electron Probe Microanalysis

Quantitative electron probe analysis is based on models based on the physics or x-ray generation, empirically adjusted to the analyses of specimens of known composition. Their accuracy can be estimated by applying them to a set of specimens of presumably well-known composition.     

Schäden durch kerbinduzierte Schwingbrüche

Failures by notch induced fatigue fractures Notches produce local stress concentrations and promote the formation of fatigue cracks. Characteristic types of notches as a result of design, manufacturing and operation are presented and remedial measures are discussed. Examples demonstrate the effect of different kinds of notches on various components.