Statistical relational learning (SRL) and graph neural networks (GNNs) are two powerful approaches for learning and inference over graphs. Typically, they are evaluated in terms of simple metrics such as accuracy over individual node labels. Complex aggregate graph queries (AGQ) involving multiple nodes, edges, and labels are common in the graph mining community and are used to estimate important network properties such as social cohesion and influence. While graph mining algorithms support AGQs, they typically do not take into account uncertainty, or when they do, make simplifying assumptions and do not build full probabilistic models. In this paper, we examine the performance of SRL and GNNs on AGQs over graphs with partially observed node labels. We show that, not surprisingly, inferring the unobserved node labels as a first step and then evaluating the queries on the fully observed graph can lead to sub-optimal estimates, and that a better approach is to compute these queries as an expectation under the joint distribution. We propose a sampling framework to tractably compute the expected values of AGQs. Motivated by the analysis of subgroup cohesion in social networks, we propose a suite of AGQs that estimate the community structure in graphs. In our empirical evaluation, we show that by estimating these queries as an expectation, SRL-based approaches yield up to a 50-fold reduction in average error when compared to existing GNN-based approaches.
The sintering properties of La1−xSrxFeO3−δ (x = 0.1, 0.25) mixed conductors have been investigated with particular emphasis on the effect of secondary phases due to cation non-stoichiometry (±5 mol% La excess and deficiency). Secondary phases, located at grain boundaries in cation non-stoichiometric materials, increased the sintering temperature compared to single-phase materials. Extensive swelling in final stage of sintering was observed in all materials, which resulted in micro-porous materials. The swelling was most pronounced in the phase pure and two-phase materials due to La-deficiency, while refractory secondary phases in La-excess materials inhibited both sintering, grain growth and swelling. In La-deficient materials, formation of molten secondary phases resulted in rapid swelling due to viscous flow. The present findings demonstrated the importance of controlling sintering temperature and time, as well as careful control of the cation stoichiometry of La1−xSrxFeO3−δ in order to achieve fully dense and homogenous La1−xSrxFeO3−δ ceramics. 相似文献
BACKGROUND: The high incidence of locoregional recurrences and distant metastases after curative surgery for gastric cancer calls for improved locoregional control and systemic adjuvant treatment. METHODS: In a randomized clinical trial on adjuvant FAM2 chemotherapy, quality of surgery was evaluated by comparing surgical and pathology data. Univariate and multivariate analysis was made to evaluate the effect of prognostic factors on survival and time of recurrence in relation to patients, tumor, and therapy. RESULTS: Of 314 patients randomized from 28 European institutions, 159 comprised the control and 155 the FAM2 group. After a median follow-up of 80 months, no statistically significant difference was found between survivals. However, for recurrence time, treated patients had a significant advantage over controls (p = 0.02). At univariate analysis, statistically significant differences in survival and time to progression emerged for T, N, disease stage and "adequacy" of surgery. The multivariate analysis retained preoperative Hb level, T, N, and "adequacy" of surgery for time of survival; and T, N, "adequacy" of surgery and adjuvant chemotherapy for recurrence time. CONCLUSIONS: Disease stage is the most important prognostic factor. "Adequate" surgery has an important effect. Adjuvant FAM2 delayed time of recurrence, but did not influence overall survival. 相似文献
Microbiological sampling of Norvegia cheese from three cheese factories was done during ripening. The evolution of aerobic mesophilic bacteria, lactococci, lactobacilli, enterococci, presumptive leuconostoc and pediococci was investigated after 30, 90, 180 and 270 days of ripening. Isolates (135) of non-starter lactic acid bacteria (NSLAB) from nine Norvegia cheeses after 90, 180 and 270 days of ripening were examined. The isolates were tested by physiological and biochemical assays, species-specific PCR and 16S rDNA sequencing. After 90 days of ripening Leuconostoc spp., most probably from the starter, and the NSLAB specie Lactobacillus paracasei dominated among the isolates, however, after longer ripening Lb. paracasei dominated. The development and evolution of the microflora in Norvegia varied according to dairy and ripening time. 相似文献
Discriminatory and robust typing methods are needed to improve the understanding of the dynamics of food-borne Campylobacter infections and epidemiology in primary animal production. To evaluate the strain discriminatory potential of typing methods, flaA short variable region (SVR) sequencing and Fourier transform infrared (FTIR) spectroscopy were applied on a collection of 102 epidemiologically related and unrelated Campylobacter jejuni strains. Previous application of FTIR spectroscopy for subtyping of Campylobacter has been limited. A subset of isolates, initially discriminated by flaA SVR sequencing, were further subjected to multilocus sequence typing (MLST). It was found that flaA SVR sequencing had a slightly higher discriminatory power than FTIR spectroscopy, based on the Simpson diversity index. The clustering of strains indicated that FTIR spectroscopy is indeed a suitable method for discrimination of Campylobacter. The isolates were assigned to six clusters based on flaA SVR sequences and nine clusters based on the FTIR spectroscopy profiles. Furthermore, the cluster analysis of flaA SVR sequences, MLST, and FTIR spectroscopy profiles showed a high degree of congruence, assigning the isolates to similar cluster structures. In conclusion, FTIR spectroscopy can be applied for subtyping of Campylobacter, and the high discriminatory potential of both flaA SVR sequencing and FTIR spectroscopy render them suitable screening methods for large numbers of strains. 相似文献