Impact of hydrated magnesium carbonate additives on the carbonation of reactive MgO cements

Reactive magnesia (MgO) cements have emerged as a potentially more sustainable and technically superior alternative to Portland cement due to their lower production temperature and ability to sequester significant quantities of CO2. Porous blocks containing MgO were found to achieve higher strength values than PC blocks. A number of variables are investigated to achieve maximum carbonation and associated high strengths. This paper focuses on the impact of four different hydrated magnesium carbonates (HMCs) as cement replacements of either 20 or 50%. Accelerated carbonation (20 C, 70-90% RH, 20% CO2) is compared with natural curing (20 C, 60-70% RH, ambient CO2). SEM, TG/DTA, XRD, and HCl acid digestion are utilized to provide a thorough understanding of the performance of MgO-cement porous blocks. The presence of HMCs resulted in the formation of larger size carbonation products with a different morphology than those in the control mix, leading to significantly enhanced carbonation and strength. © 2013 Elsevier Ltd.

Avoiding pathologies in very deep networks

Choosing appropriate architectures and regularization strategies of deep networks is crucial to good predictive performance. To shed light on this problem, we analyze the analogous problem of constructing useful priors on compositions of functions. Specifically, we study the deep Gaussian process, a type of infinitely-wide, deep neural network. We show that in standard architectures, the representational capacity of the network tends to capture fewer degrees of freedom as the number of layers increases, retaining only a single degree of freedom in the limit. We propose an alternate network architecture which does not suffer from this pathology. We also examine deep covariance functions, obtained by composing infinitely many feature transforms. Lastly, we characterize the class of models obtained by performing dropout on Gaussian processes.