Search

Identification of the SARS-CoV-2 virus by RT-PCR from a nasopharyngeal swab sample is a common test for diagnosing COVID-19. However, some patients present clinical, laboratorial, and radiological evidence of COVID-19 infection with negative RT-PCR result(s). Thus, we assessed whether positive results were associated with intubation and mortality. This study was conducted in a Brazilian tertiary hospital from March to August of 2020. All patients had clinical, laboratory, and radiological diagnosis of COVID-19. They were divided into two groups: positive (+) RT-PCR group, with 2292 participants, and negative (−) RT-PCR group, with 706 participants. Patients with negative RT-PCR testing and an alternative most probable diagnosis were excluded from the study. The RT-PCR(+) group presented increased risk of intensive care unit (ICU) admission, mechanical ventilation, length of hospital stay, and 28-day mortality, when compared to the RT-PCR(−) group. A positive SARS-CoV-2 RT-PCR result was independently associated with intubation and 28 day in-hospital mortality. Accordingly, we concluded that patients with a COVID-19 diagnosis based on clinical data, despite a negative RT-PCR test from nasopharyngeal samples, presented more favorable outcomes than patients with positive RT-PCR test(s).

Recent advances in applying deep learning to planning have shown that Deep Reactive Policies (DRPs) can be powerful for fast decision-making in complex environments. However, an important limitation of current DRP-based approaches is either the need of optimal planners to be used as ground truth in a supervised learning setting or the sample complexity of high-variance policy gradient estimators, which are particularly troublesome in continuous state-action domains. In order to overcome those limitations, we introduce a framework for training DRPs in continuous stochastic spaces via gradient-based policy search. The general approach is to explicitly encode a parametric policy as a deep neural network, and to formulate the probabilistic planning problem as an optimization task in a stochastic computation graph by exploiting the re-parameterization of the transition probability densities; the optimization is then solved by leveraging gradient descent algorithms that are able to handle non-convex objective functions. We benchmark our approach against stochastic planning domains exhibiting arbitrary differentiable nonlinear transition and cost functions (e.g., Reservoir Control, HVAC and Navigation). Results show that DRPs with more than 125,000 continuous action parameters can be optimized by our approach for problems with 30 state fluents and 30 action fluents on inexpensive hardware under 6 minutes. Also, we observed a speedup of 5 orders of magnitude in the average inference time per decision step of DRPs when compared to other state-of-the-art online gradient-based planners when the same level of solution quality is required.

A novel chitosan/kaolin composite membrane with high flux and enhanced hydrophilicity was prepared by solvent casting and evaporation process in the presence of polyethylene glycol (PEG). The characterization of the as-fabricated membranes indicated that the combined effect of kaolin as reinforcing agent and PEG as porogen into a chitosan matrix showed significant influence on the morphology and hydrophilicity. Highly porous membranes with finger-like structure were obtained. Both size and density of the pores increased with increasing the PEG content from 0 to 3 wt%. The incorporation of the hydrophilic polymeric additive constructed a water channel network and increased the free volume within the composite matrix, contributing to the enhanced filtration performance of the resulting membrane. Additionally, their mechanical, thermal and chemical stability showed a large improvement compared to the unmodified one.

This paper describes a real application of graphical neural networks (GNNs) in the dynamic estimation of spatially distributed buoys that are of central importance in maritime navigation. We describe the techniques we used to process both data and background knowledge about the domain, indicating why GNNs are particularly well suited for this sort of task. We report our empirical results, demonstrating that GNNs profitably use the avaible relational structure.

Probabilistic circuits are deep probabilistic models with neural-network-like semantics capable of accurately and efficiently answering probabilistic queries without sacrificing expressiveness. Probabilistic Sentential Decision Diagrams (PSDDs) are a subclass of probabilistic circuits able to embed logical constraints to the circuit’s structure. In doing so, they obtain extra expressiveness with empirical optimal performance. Despite achieving competitive performance compared to other state-of-the-art competitors, there have been very few attempts at learning PSDDs from a combination of both data and knowledge in the form of logical formulae. Our work investigates sampling random PSDDs consistent with domain knowledge and evaluating against state-of-the-art probabilistic models. We propose a method of sampling that retains important structural constraints on the circuit’s graph that guarantee query tractability. Finally, we show that these samples are able to achieve competitive performance even on larger domains.

Probabilistic sentential decision diagrams are logic circuits where the inputs of disjunctive gates are annotated by probability values. They allow for a compact representation of joint probability mass functions defined over sets of Boolean variables, that are also consistent with the logical constraints defined by the circuit. The probabilities in such a model are usually learned from a set of observations. This leads to overconfident and prior-dependent inferences when data are scarce, unreliable or conflicting. In this work, we develop the credal sentential decision diagrams, a generalisation of their probabilistic counterpart that allows for replacing the local probabilities with (so-called credal) sets of mass functions. These models induce a joint credal set over the set of Boolean variables, that sharply assigns probability zero to states inconsistent with the logical constraints. Three inference algorithms are derived for these models, these allow to compute: (i) the lower and upper probabilities of an observation for an arbitrary number of variables; (ii) the lower and upper conditional probabilities for the state of a single variable given an observation; (iii) whether or not all the probabilistic sentential decision diagrams compatible with the credal specification have the same most probable explanation of a given set of variables given an observation of the other variables. These inferences are tractable, as all the three algorithms, based on bottom-up traversal with local linear programming tasks on the disjunctive gates, can be solved in polynomial time with respect to the circuit size. For a first empirical validation, we consider a simple application based on noisy seven-segment display images. The credal models are observed to properly distinguish between easy and hard-to-detect instances and outperform other generative models not able to cope with logical constraints., To appear in the International Journal of Approximate Reasoning (IJAR Volume 125)

Poly(rotaxane)s were synthesised by capping with 2,4-dinitrofluorobenzene the corresponding pseudopoly(rotaxane)s consisting of α-cyclodextrin (α-CD) threaded onto two end-aminated poly(ethyleneglycol) (PEG, Mw 2000 and 3500), with the aim to be used for nanoscale building blocks of facilitated transport membranes. The number of α-CD threaded in poly(rotaxane)s obtained proved to remain constant whatever the initial molar ratio ethylene oxide units: α-CD under our conditions of capping. The resulting polymers had a similar polymolecularity index as the parent PEGs. Thermal properties and light scattering analysis in DMSO showed that inclusion complexation in α-CD stiffens the PEG chains rendering them as semi rod-like macromolecules. A strong tendency to auto-association was observed which is attributed to hydrophobic interactions between the blocking end-groups. Films were prepared by crosslinking poly(vinylalcohol) (PVA) and poly(rotaxane)s with hexamethylenediisocyanate. After removing the end-stoppers under mild alkaline conditions and releasing PEG from the CD cavities, membranes were studied in a water/methanol pertraction system. Large increases in the permeation rate of toluene were observed by replacing α-CD with an α-CD molecular assembly. This effect was interpreted by the cooperativity of CD cavities in the toluene permeation across channels formed inside the membrane.

International audience; In response to the fresh water scarcity, Tunisia is utilizing more and more membrane desalination of unconventional resources, including brackish waters and seawater. The widespread reserves of groundwaters and their low salinity make this resource of special interest. Two predominant ionic compositions have been identified depending on their relative proportion of sulfate to chloride ions. The question arising for the decision-makers concerns the choice of membrane technology and, therefore, of membrane. Two nanofiltration (NF) membranes (NF270 and NF90) and a reverse osmosis (RO) one (BW30) were tested in a desalination study of synthetic feeds reproducing the ionic composition of three representative groundwaters. Sulfate/chloride ratio appears to be the key factor for the membrane choice to obtain good quality drinking water meeting the Tunisian standards. Moreover, validation of two prediction tools was investigated: ROSA, software provided by the membrane manufacturer and Nanoflux ® , software specifically designed for NF. The experimental NF results are well fitted by the Nanoflux ® simulations. We concluded that ROSA cannot generally provide good NF predictions because it does not take into account the electric interactions between membrane and feed.

An important and often neglected aspect in probabilistic planning is how to account for different attitudes towards risk in the process. In goal-driven problems, modeled as Shortest Stochastic Path (ssp) problems, risk arises from the uncertainties on future events and how they can lead to goal states. An ssp agent that minimizes the expected accumulated cost is considered a risk-neutral agent, while with a different optimization criterion it could choose between two extreme attitudes: risk-aversion or risk-prone. In this work we consider a Risk Sensitive ssp (called rs-ssp) that uses an expected exponential utility parameterized by the risk factor \(\lambda \) that is used to define the agent’s risk attitude. Moreover, a \(\lambda \)-value is feasible if it admits a policy with finite expected cost. There are several algorithms capable of determining an optimal policy for rs-ssp s when we fix a feasible value for \(\lambda \). However, so far, there has been only one approach to find an extreme \(\lambda \) feasible i.e., an extreme risk-averse policy. In this work we propose and compare new approaches to finding the extreme feasible \(\lambda \) value for a given rs-ssp, and to return the corresponding extreme risk-averse policy. Experiments on three benchmark domains show that our proposals outperform previous approach, allowing the solution of larger problems.

Planning problems with continuous state and action spaces are difficult to solve with existing planning techniques, specially when the state transition is defined by a high-dimension non-linear dynamics. Recently, a technique called Planning through Backpropagation (PtB) was introduced as an efficient and scalable alternative to traditional optimization-based methods for continuous planning problems. PtB leverages modern gradient descent algorithms and highly optimized automatic differentiation libraries to obtain approximate solutions. However, to date there have been no empirical evaluations comparing PtB with Linear-Quadratic (LQ) control problems. In this work, we compare PtB with an optimal algorithm from control theory called LQR, and its iterative version iLQR, when solving linear and non-linear continuous deterministic planning problems. The empirical results suggest that PtB can be an efficient alternative to optimizing non-linear continuous deterministic planning, being much easier to be implemented and stabilized than classical model-predictive control methods.

We analyze the computational complexity of probabilistic logic programming with constraints, disjunctive heads, and aggregates such as sum and max. We consider propositional programs and relational programs with bounded-arity predicates, and look at cautious reasoning (i.e., computing the smallest probability of an atom over all probability models), cautious explanation (i.e., finding an interpretation that maximizes the lower probability of evidence) and cautious maximum-a-posteriori (i.e., finding a partial interpretation for a set of atoms that maximizes their lower probability conditional on evidence) under Lukasiewicz's credal semantics.

International audience; This work is aimed at developing a new process to separate alpha-cyclodextrin (CD) from binary alpha-CD/beta-CD mixtures. To achieve this objective, a membrane process was designed with water as solvent and using the selective accommodation of a guest Igepal Co-890 (IG) within the beta-CD cavity. In the first part, host-guest inclusion study suggested the formation of IG(2)/beta-CD complex in presence of IG excess. Moreover, increasing IG/CD molar ratio to 1.5 leads to the complete complexation of beta-CD while alpha-CD remains free. The second part was devoted to the separation of alpha- and beta-CD using UF membranes (GH and GK from GE-Osmonics). The influence of parameters including membrane molecular weight cut-off (MWCO), transmembrane pressure (Delta P), IG/beta-CD and alpha-CD/beta-CD molar ratios, was investigated. MWCO, Delta P, IG/beta-CD and alpha-CD/beta-CD molar ratios respectively of 1000 Da, 2 bar, 1.5 and 11:89 are the optimized values to reach an alpha-CD separation factor as high as 38. The experimental data fits very well with predicted values calculated using the parameters found in the complexation study and rejection of single solutions indicating that selectivity predominantly originates from steric exclusion of species by membrane GH.

International audience; Poly(ether imide) (PEI) membrane with enhanced antifouling property was successfully prepared in a mild and simple procedure. The virgin membrane was firstly functionalized with an aqueous solution of diamino-terminated poly(ethylene oxide) block copolymer (PEG-diamine). Glutaraldehyde was used in a second step as a linker to chemically attach additional PEG-diamine to the primary amine groups grafted on PEI membrane surface. Immobilization of PEG segments was confirmed using Fourier transform infrared spectroscopy, X-ray photoelectron spectroscopy, scanning electron microscopy and atomic force microscopy. Ultrafiltration experiments revealed that the enhancement of a PEG coverage on the membrane surface provided superior anti-protein-fouling property. Cycles of protein filtration also demonstrated that the antifouling surface was stable over time and excellent ultrafiltration performance could be maintained without the need of harsh cleansing operation.

Editorial

Sweeping gas membrane distillation (SGMD) is a useful option for dehydration of aqueous solvent solutions. This study investigated the technical viability and competitiveness of the use of SGMD to concentrate aqueous solutions of 1,3-dimethyl-2-imidazolidinone (DMI), a dipolar aprotic solvent. The concentration from 30% to 50% of aqueous DMI solutions was attained in a bench installation with Liqui-Cel SuperPhobic&reg, hollow-fiber membranes. The selected membranes resulted in low vapor flux (below 0.15 kg/h&middot, m2) but were also effective for minimization of DMI losses through the membranes, since these losses were maintained below 1% of the evaporated water flux. This fact implied that more than 99.2% of the DMI fed to the system was recovered in the produced concentrated solution. The influence of temperature and flowrate of the feed and sweep gas streams was analyzed to develop simple empirical models that represented the vapor permeation and DMI losses through the hollow-fiber membranes. The proposed models were successfully applied to the scaling-up of the process with a preliminary multi-objective optimization of the process based on the simultaneous minimization of the total membrane area, the heat requirement and the air consumption. Maximal feed temperature and air flowrate (and the corresponding high operation costs) were optimal conditions, but the excessive membrane area required implied an uncompetitive alternative for direct industrial application.

Recent research has shown the potential of learning lane following policies from annotated video sequences through the use of advanced machine learning techniques. They however require high computational power, prohibiting their use in low-budget projects such as educational robotic kits and embedded devices. Sum-product networks (SPNs) are a class of deep probabilistic models with clear probabilistic semantics and competitive performance. Importantly, SPNs learned from data are usually several times smaller than deep neural networks trained for the same task. In this work, we develop an end-to-end imitation learning solution to lane following using SPNs to classify images into a finite set of actions. Images are obtained from a monocular camera, which is part of the low-cost custom made mobile robot. Our results show that our solution generalizes training conditions with relatively few data. We investigate the trade-off between computational and predictive performance, and conclude that sacrificing accuracy for the benefit of faster inference results in improved performance in the real world, especially in resource constrained environments.

The DST/Mintek National Innovation Center, the Institute of Nanotechnology and Water – Water Node at the University of Johannesburg and the UNESCO SIMEV Chair, hosted by the European Institute of Membranes, joined hands to organize from 17 to 19 July 2019 the first workshop on membrane development and separation processes for water and air purification in South Africa. This booklet of abstracts gathers the detailed program with the contacts of all the speakers, with their short biography and abstracts.

International audience; AB(2)-type amphiphilic (HPMC)(2)-PLA copolymers with various hydrophilic block lengths were synthesized using a three step procedure: ring-opening polymerization of L-lactide initiated by propynol, amination reduction of the aldehyde endgroup of HPMC, and thiol-click reaction. The resulted copolymers were characterized by NMR, DOSY-NMR, SEC and FT-IR. The cloud point (CP) was determined by UV-vis spectrometer. Data show that the HPMC block length has little effect on the Cp of the copolymers which is lower than that of HPMC. The self-assembly behavior of the copolymers was investigated from DLS, TEM, and critical micelle concentration (CMC) measurements. Spherical micelles are obtained by self-assembly of copolymers in aqueous solution. The micelle size and the CMC of copolymers increase with increasing HPMC block length. It is concluded that biobased and biodegradable (HPMC) 2-PLA copolymers could be promising as nano-carrier of hydrophobic drugs.