Secondary Organic Aerosol from OH-Initiated Oxidation of Mixtures of d -Limonene and β-Myrcene
Not Available
Engineering topics
Publications and source records attributed to Liu, Sijia.
Not Available
K-Nearest Neighbor (kNN)-based deep learning methods have been applied to many applications due to their simplicity and geometric interpretability. However, the robustness of kNN-based deep classification models has not been thoroughly explored and kNN attack strategies are underdeveloped. In this paper, we first propose an Adversarial Soft kNN (ASK) loss for developing more effective kNN-based deep neural network attack strategies and designing better defense methods against them. Our ASK loss provides a differentiable surrogate of the expected kNN classification error. It is also interpretable as it preserves the mutual information between the perturbed input and the in-class-reference data. We use the ASK loss to design a novel attack method called the ASK-Attack (ASK-Atk), which shows superior attack efficiency and accuracy degradation relative to previous kNN attacks on hidden layers. We then derive an ASK-Defense (ASK-Def) method that optimizes the worst-case ASK training loss. Experiments on CIFAR-10 (ImageNet) show that (i) ASK-Atk achieves ≥13% (≥ 13% ) improvement in attack success rate over previous kNN attacks, and (ii) ASK-Def outperforms the conventional adversarial training method by ≥ 6.9% (≥ 3.5% ) in terms of robustness improvement. Relevant codes are available at https://github.com/wangren09/ASK .
Explore the source record for details and available documents.
Heavy metals (HMs) have become a major environmental pollutant threatening ecosystems and human health. Although hyperaccumulators provide a viable alternative for the bioremediation of HMs, the potential of phytoremediation is often limited by the small biomass and slow growth rate of hyperaccumulators and HM toxicity to plants. Here, plant growth-promoting bacteria (PGPB)-assisted phytoremediation was used to enhance the phytoremediation of HM-contaminated soils. A PGPB with HM-tolerant (HMT-PGPB), Bacillus sp. PGP15 was isolated from the rhizosphere of a cadmium (Cd) hyperaccumulator, Solanum nigrum . Pot experiments demonstrated that inoculation with strain PGP15 could significantly increase the growth of S. nigrum . More importantly, strain PGP15 markedly improved Cd accumulation in S. nigrum while alleviating Cd-induced stress in S. nigrum . Specifically, PGP15 inoculation significantly decreased the contents of H 2 O 2 , MDA, and O 2 · − in S. nigrum , while the activities (per gram plant fresh weight) of SOD, APX, and CAT were significantly increased in the PGP15-inoculated plants compared with the control sample. These results suggested that the interactions between strain PGP15 and S. nigrum could overcome the limits of phytoremediation alone and highlighted the promising application potential of the PGPB-hyperaccumulator collaborative pattern in the bioremediation of HM-contaminated soils. Furthermore, the PGP15 genome was sequenced and compared with other strains to explore the mechanisms underlying plant growth promotion by HMT-PGPB. The results showed that core genes that define the fundamental metabolic capabilities of strain PGP15 might not be necessary for plant growth promotion. Meanwhile, PGP15-specific genes, including many transposable elements, played a crucial role in the adaptive evolution of HM resistance. Overall, our results improve the understanding of interactions between HMT-PGPB and plants and facilitate the application of HMT-PGPB in the phytoremediation of HM-contaminated soils.
Here, the ability to prepare and tailor synthetic analogues of native bacteriochlorophylls enables diverse applications. A de novo route entails dimerization of a dihydrodipyrrin-acetal to afford the corresponding 5-methoxy and/or 5-unsubstituted bacteriochlorin, wherein each pyrroline ring contains a gem-dimethyl group to ensure stability toward adventitious dehydrogenation. The presence of a 5-methoxy group facilitates bromination at the distal meso-(15-)position. While bromination of 5-unsubstituted bacteriochlorins typically affords a mixture of brominated products, here the presence of two substitution patterns (2,12-dicarboethoxy, 2,12-diacetyl) has been found to facilitate selective meso-bromination in the absence of the methoxy substituent. The introduction of a single meso-bromine atom in a bacteriochlorin opens opportunities for Pd-mediated derivatization, which include (1) preparation of four ethynylphenyl building blocks (and two benchmark bacteriochlorins) with long-wavelength absorption bands tuned across 725–757 nm, for use in preparation of multichromophore arrays; (2) installation of a bioconjugatable group to free base bacteriochlorins or a copper bacteriochlorin, the latter for possible use in photoacoustic imaging; and (3) installation of an S-acetylthio group for surface attachment. Altogether, 25 new bacteriochlorins are described including 5 meso-bromobacteriochlorin intermediates and 12 target bacteriochlorins.
Synthetic bacteriochlorins, analogues of native bacteriochlorophylls, provide absorption in the near-infrared spectral region and hence are valuable for fundamental studies in photochemistry and applications ranging from solar energy conversion to photomedicine. The full realization of such opportunities hinges on access to synthetically malleable bacteriochlorins. Here, two established routes that rely on self-condensation of a dihydrodipyrrin-acetal or dihydrodipyrrin-carboxaldehyde have been exploited to prepare 6 bacteriochlorins (2 new, 4 known in improved fashion). Each bacteriochlorin contains a gem-dimethyl group in the reduced, pyrroline ring to ensure stability toward adventitious dehydrogenation. These and related synthetic bacteriochlorins have been derivatized to tailor the perimeter of the macrocycle with diverse groups, affording 9 additional new bacteriochlorins and 4 known bacteriochlorins. By direct installation and/or derivatization, the scope of attached entities in this collection includes alkyl, amino, aryl, bromo, carboethoxy, oxo, phenylethynyl, and pyridyl. Furthermore, the bacteriochlorins are suited for surface attachment, bioconjugation, water-solubilization, vibrational studies, and elaboration into multichromophore arrays. Altogether the synthesis of 25 new compounds (11 bacteriochlorins, 14 precursors) is described.
Solubilization of hydrophobic tetrapyrrole macrocycles and other fluorophores in aqueous solution has been achieved by covalent attachment to the terminus of an amphiphilic polymer, thereby affording a single-fluorophore–single-chain nanoparticle (SCNP). The polymer is a heterotelechelic random polyacrylate/polyacrylamide copolymer bearing hydrophobic and hydrophilic pendant groups. The polymer has a folded unimeric morphology (~13 nm hydrodynamic diameter) in 1 M NaCl aqueous solution as indicated by dynamic light-scattering spectroscopy. Five hydrophobic organic fluorophores (coumarin, perylene, two chlorins, one phthalocyanine) have been synthesized with a conjugatable tether. Covalent conjugation of the fluorophore–maleimide to the polymer terminus via thiol–maleimide reaction was carried out in DMF. The resulting fluorophore–SCNP in 1 M NaCl aqueous solution exhibited retained spectral features and fluorescence quantum yield comparable to those of the respective hydrophobic benchmark in toluene. This single-fluorophore–single-polymer strategy simplifies the challenging and often idiosyncratic syntheses of water-soluble tetrapyrrole macrocycles by using the polymer as a general aqueous solubilization package, and in so doing opens up opportunities for the application of hydrophobic fluorophores in the life sciences.
We study the most practical problem setup for evaluating adversarial robustness of a machine learning system with limited access: the hard-label black-box attack setting for generating adversarial examples, where limited model queries are allowed and only the decision is provided to a queried data input. Several algorithms have been proposed for this problem but they typically require huge amount (>20,000) of queries for attacking one example. Among them, one of the state-of-the-art approaches (Cheng et al., 2019) showed that hard-label attack can be modeled as an optimization problem where the objective function can be evaluated by binary search with additional model queries, thereby a zeroth order optimization algorithm can be applied. In this paper, we adopt the same optimization formulation but propose to directly estimate the sign of gradient at any direction instead of the gradient itself, which enjoys the benefit of single query. Using this single query oracle for retrieving sign of directional derivative, we develop a novel query-efficient Sign-OPT approach for hard-label black-box attack. We provide a convergence analysis of the new algorithm and conduct experiments on several models on MNIST, CIFAR-10 and ImageNet. We find that Sign-OPT attack consistently requires 5X to 10X fewer queries when compared to the current state-of-the-art approaches, and usually converges to an adversarial example with smaller perturbation.