Search

I propose that meta-learned models, and in particular the situation-aware deployment of "learning-to-infer" modules can be advantageously extended to domains commonly thought to lie outside the cognitive, such as motivations and preferences on one hand, and the effectuation of micro- and coping-type behaviors.

I argue for a more complicated but nonetheless computationally feasible and algorithmically intelligible interplay between exploration and exploitation and for admitting into our conceptual toolkit regimes of exploitative exploration and exploratory exploitation that can enhance the novelty and usefulness of the results of either problemistic or serendipitous search.

I consider the dynamics of regulator-agent interactions in situations in which there are significant mismatches in their abilities to discern and register information and calculate and act upon successful inferences.

Depending on what we mean by "explanation," challenges to the explanatory depth and reach of deep neural network models of visual and other forms of intelligent behavior may need revisions to both the elementary building blocks of neural nets (the explananda) and to the ways in which experimental environments and training protocols are engineered (the explanantia). The two paths assume and imply sharply different conceptions of how an explanation explains and of the explanatory function of models.

Can we explain the advantage natural languages enjoy over ideographies in a way that enables us to attempt the design of an ideography that "works"? I deploy an adapted version of Shannon's source- and channel-coding partitioning of a communication system to explain the communicative dynamics and shortfalls of ideographies, and reveal ways in which entrenchable, generalist ideographies could be designed.

In this paper, we study distributed inference and learning over networks which can be modeled by a directed graph. A subset of the nodes observes different features, which are all relevant/required for the inference task that needs to be performed at some distant end (fusion) node. We develop a learning algorithm and an architecture that can combine the information from the observed distributed features, using the processing units available across the networks. In particular, we employ information-theoretic tools to analyze how inference propagates and fuses across a network. Based on the insights gained from this analysis, we derive a loss function that effectively balances the model's performance with the amount of information transmitted across the network. We study the design criterion of our proposed architecture and its bandwidth requirements. Furthermore, we discuss implementation aspects using neural networks in typical wireless radio access and provide experiments that illustrate benefits over state-of-the-art techniques.

Conducting hydromorphological assessments for evaluating the ecological status of rivers is a key requirement of the Directive 2000/60/EC (Water Framework Directive - WFD) within European Union (EU) Member States. This paper aims at understanding how this requirement was implemented, through an original comparative review of methodologies for rivers' hydromorphological assessment in three EU Member States, which joined the EU at different times, and with many differences in terms of hydrographic features, socio-economic and water management systems: France, Romania, and Croatia. More precisely, the paper aims at identifying and understanding the main principles guiding the hydromorphological assessment methodologies, elements and data used, giving an overview of the results of hydromorphological river status assessment, and concluding on the stage of hydromorphological assessment implementation. France developed numerous methodologies for physical habitat survey since the 1990s and it is currently conducting a rigorous hydromorphological field survey, but it does not yet have any national methodology for rivers' hydromorphological status assessment, nevertheless foreseen for the next cycle of the WFD. Besides, Romania and Croatia have already started the assessment of the hydromorphological status of rivers within the two cycles of the River Basin Management Plans and are making efforts to improve the hydromorphological monitoring activity. The methods generally differ in indicators, data used, and spatial scale of analysis, which makes it difficult to compare the results of the assessments. Despite a common water policy, the methodological dissimilarities seem to be rather usual between EU Member States. Therefore, the standardization of methodologies appears to be necessary, but the current results could be useful for setting priorities for river restoration and for achieving a better status at a national scale., (Copyright © 2018 Elsevier Ltd. All rights reserved.)

Understanding how microbial diversity influences ecosystem properties is of paramount importance. Cellular traits-which determine responses to the abiotic and biotic environment-may help us rigorously link them. However, our capacity to measure traits in natural communities has thus far been limited. Here we compared the predictive power of trait richness (trait space coverage), evenness (regularity in trait distribution) and divergence (prevalence of extreme phenotypes) derived from individual-based measurements with two species-level metrics (taxonomic richness and evenness) when modelling the productivity of natural phytoplankton communities. Using phytoplankton data obtained from 28 lakes sampled at different spatial and temporal scales, we found that the diversity in individual-level morphophysiological traits strongly improved our ability to predict community resource-use and biomass yield. Trait evenness-the regularity in distribution of individual cells/colonies within the trait space-was the strongest predictor, exhibiting a robust negative relationship across scales. Our study suggests that quantifying individual microbial phenotypes in trait space may help us understand how to link physiology to ecosystem-scale processes. Elucidating the mechanisms scaling individual-level trait variation to microbial community dynamics could there improve our ability to forecast changes in ecosystem properties across environmental gradients.