Water serves as an inert environment for the dispersion and application of many kinds of herbicides. Viologen compounds, a type of widely used but highly toxic herbicide, are stable in bulk water, whose half-life can be up to 23 weeks in natural water, imposing a severe health risk to mammals. In this study, we present the striking results of the spontaneous and ultrafast reduction-induced degradation of three viologen compounds in water microdroplets and provide the concentration, time, temperature dependence, mechanism, and scale-up of the reactions. We postulate that the electrons existing at the air−water interface of the microdroplets due to the unique redox potential therein initiate the reduction, from which further degradation occurs. The host−guest complexation between cucurbit[7]uril and viologens only slightly changes the redox potential of viologens in the bulk but completely inhibits the reactions in microdroplets, adding to the uniqueness of the redox potentials at the air−water interfaces of microdroplets. Taken together, microdroplets might have been functioning as naturally occurring ubiquitous tiny electrochemical cells for a plethora of unique redox reactions that were thought to be impossible in the bulk water.

No abstract

No abstract

Addressing catastrophic forgetting is one of the key challenges in continual learning where machine learning systems are trained with sequential or streaming tasks. Despite recent remarkable progress in state-of-the-art deep learning, deep neural networks (DNNs) are still plagued with the catastrophic forgetting problem. This paper presents a conceptually simple yet general and effective framework for handling catastrophic forgetting in continual learning with DNNs. The proposed method consists of two components: a neural structure optimization component and a parameter learning and/or fine-tuning component. By separating the explicit neural structure learning and the parameter estimation, not only is the proposed method capable of evolving neural structures in an intuitively meaningful way, but also shows strong capabilities of alleviating catastrophic forgetting in experiments. Furthermore, the proposed method outperforms all other baselines on the permuted MNIST dataset, the split CIFAR100 dataset and the Visual Domain Decathlon dataset in continual learning setting.