Interpreting subjectivity in causal relations takes effort: Subjective, claim-argument relations are read slower than objective, cause-consequence relations. In an eye-tracking-while-reading experiment, we investigated whether connectives and stance markers can play a facilitative role. Sixty-five Chinese participants read sentences expressing a subjective causal relation, systematically varied in the use of stance markers (no, attitudinal, epistemic) in the first clause and connectives (neutral suoyi “so”, subjective kejian “so”) in the second clause. Results showed that processing subjectivity proceeds highly incrementally: The interplay of the subjectivity markers is visible as the sentence unfolds. Subjective connectives increased reading times, irrespective of the type of stance marker being used. Stance markers did, however, facilitate the processing of modal verbs in subjective relations. We conclude that processing subjectivity involves evaluating how the argument supports the claim and that connectives, modal verbs, and stance markers function as processing instructions that help readers achieve this evaluation.
Artificial neural networks (ANNs), like convolutional neural networks (CNNs), have achieved the state-of-the-art results formanymachine learning tasks. However, inference with large-scale full-precision CNNs must cause substantial energy consumption and memory occupation, which seriously hinders their deployment on mobile and embeddedsystems. Highly inspired from biological brain, spiking neural networks (SNNs) are emerging as new solutions because of natural superiority in brain-like learning and great energy efficiency with event-driven communication and computation. Nevertheless, training a deep SNN remains a main challenge and there is usually a big accuracy gap between ANNs and SNNs. In this paper, we introduce a hardware-friendly conversion algorithm called “scatter-and-gather” to convert quantized ANNs to lossless SNNs, where neurons are connected with ternary {−1,0,1} synaptic weights. Each spiking neuron is stateless and more like original McCulloch and Pitts model, because it fires at most one spike and need be reset at each time step. Furthermore, we develop an incremental mapping framework to demonstrate efficient network deployments on a reconfigurable neuromorphic chip. Experimental results show our spiking LeNet on MNIST and VGG-Net on CIFAR-10 datasetobtain 99.37% and 91.91% classification accuracy, respectively. Besides, the presented mapping algorithm manages network deployment on our neuromorphic chip with maximum resource efficiency and excellent flexibility. Our four-spike LeNet and VGG-Net on chip can achieve respective real-time inference speed of 0.38 ms/image, 3.24 ms/image, and an average power consumption of 0.28 mJ/image and 2.3 mJ/image at 0.9 V, 252 MHz, which is nearly two orders of magnitude more efficient than traditional GPUs.
Cooking has been proven to be a significant source of primary organic aerosol, especially in megacities. However, the formation of secondary organic aerosol (SOA) derived from cooking emissions is still poorly understood. In this work, four prevalent Chinese domestic cooking types involving complicated cuisines and various cooking methods were chosen to conduct a lab simulation for SOA formation using a Gothenburg potential aerosol mass reactor (Go: PAM). After samples had been aged under OH exposures of 4.3–27.1 × 1010 molecules cm–3 s, the domestic cooking SOA was characterized by mass growth potentialities (1.81–3.16), elemental ratios (O/C = 0.29–0.41), and mass spectra. Compared with other organic aerosol (OA), domestic cooking SOA is a kind of less oxidized oxygenated OA (LO-OOA) with a unique oxidation pathway (alcohol/peroxide pathway) and mass spectra (characteristic peaks at m/z 28, 29, 41, 43, 44, 55, and 57). This study is expected to identify the cooking SOA under actual cooking conditions, which could contribute to the formulation of pollution source control as well as the health risk assessment of exposure to cooking fumes.