Thousands of conductance levels in memristors integrated on CMOS.

Neural networks based on memristive devices1-3 have the ability to improve throughput and energy efficiency for machine learning4,5 and artificial intelligence6, especially in edge applications7-21. Because training a neural network model from scratch is costly in terms of hardware resources, time and energy, it is impractical to do it individually on billions of memristive neural networks distributed at the edge. A practical approach would be to download the synaptic weights obtained from the cloud training and program them directly into memristors for the commercialization of edge applications. Some post-tuning in memristor conductance could be done afterwards or during applications to adapt to specific situations. Therefore, in neural network applications, memristors require high-precision programmability to guarantee uniform and accurate performance across a large number of memristive networks22-28. This requires many distinguishable conductance levels on each memristive device, not only laboratory-made devices but also devices fabricated in factories. Analog memristors with many conductance states also benefit other applications, such as neural network training, scientific computing and even 'mortal computing'25,29,30. Here we report 2,048 conductance levels achieved with memristors in fully integrated chips with 256 × 256 memristor arrays monolithically integrated on complementary metal-oxide-semiconductor (CMOS) circuits in a commercial foundry. We have identified the underlying physics that previously limited the number of conductance levels that could be achieved in memristors and developed electrical operation protocols to avoid such limitations. These results provide insights into the fundamental understanding of the microscopic picture of memristive switching as well as approaches to enable high-precision memristors for various applications. Fig. 1 HIGH-PRECISION MEMRISTOR FOR NEUROMORPHIC COMPUTING.: a, Proposed scheme of the large-scale application of memristive neural networks for edge computing. Neural network training is performed in the cloud. The obtained weights are downloaded and accurately programmed into a massive number of memristor arrays distributed at the edge, which imposes high-precision requirements on memristive devices. b, An eight-inch wafer with memristors fabricated by a commercial semiconductor manufacturer. c, High-resolution transmission electron microscopy image of the cross-section view of a memristor. Pt and Ta serve as the bottom electrode (BE) and top electrode (TE), respectively. Scale bars, 1 µm and 100 nm (inset). d, Magnification of the memristor material stack. Scale bar, 5 nm. e, As-programmed (blue) and after-denoising (red) currents of a memristor are read by a constant voltage (0.2 V). The denoising process eliminated the large-amplitude RTN observed in the as-programmed state (see Methods). f, Magnification of three nearest-neighbour states after denoising. The current of each state was read by a constant voltage (0.2 V). No large-amplitude RTN was observed, and all of the states can be clearly distinguished. g, An individual memristor on the chip was tuned into 2,048 resistance levels by high-resolution off-chip driving circuitry, and each resistance level was read by a d.c. voltage sweeping from 0 to 0.2 V. The target resistance was set from 50 µS to 4,144 µS with a 2-µS interval between neighbouring levels. All readings at 0.2 V are less than 1 µS from the target conductance. Bottom inset, magnification of the resistance levels. Top inset, experimental results of an entire 256 × 256 array programmed by its 6-bit on-chip circuitry into 64 32 × 32 blocks, and each block is programmed into one of the 64 conductance levels. Each of the 256 × 256 memristors has been previously switched over one million cycles, demonstrating the high endurance and robustness of the devices.

High expression of RPL27A predicts poor prognosis in patients with hepatocellular carcinoma.

BACKGROUND: Hepatocellular carcinoma (HCC) is one of the most common cancers in the digestive system with rapid progression and poor prognosis. Recent studies have shown that RPL27A could be used as a biomarker for a variety of cancers, but its role in HCC is not clear. METHOD: We analyzed the expression of RPL27A in the pan-cancer analysis and analyzed the relationship between the expression of RPL27A and the clinical features and prognosis of patients with HCC. We evaluated the expression difference of RPL27A in HCC tissues and paired normal adjacent tissues using immunohistochemistry. Furthermore, we analyzed the co-expression genes of RPL27A and used them to explore the possible mechanism of RPL27A and screen hub genes effecting HCC. In addition, we studied the role of RPL27A in immune infiltration and mutation. RESULTS: We found that the expression level of RPL27A increased in a variety of cancers, including HCC. In HCC patients, the high expression of RPL27A was related to progression and poor prognosis as an independent predictor. We also constructed a protein interaction network through co-expression gene analysis of RPL27A and screened 9 hub genes. Enrichment analysis showed that co-expression genes were associated with ribosome pathway, viral replication, nuclear-transcribed mRNA catabolic process, and nonsense-mediated decay. We found that the expression level of RPL27A was closely related to TP53 mutation and immune infiltration in HCC. CONCLUSION: RPL27A might become a biomarker in the diagnosis, treatment, and follow-up of patients with HCC.

Data-driven contact network models of COVID-19 reveal trade-offs between costs and infections for optimal local containment policies.

While several non-pharmacological measures have been implemented for a few months in an effort to slow the coronavirus disease (COVID-19) pandemic in the United States, the disease remains a danger in a number of counties as restrictions are lifted to revive the economy. Making a trade-off between economic recovery and infection control is a major challenge confronting many hard-hit counties. Understanding the transmission process and quantifying the costs of local policies are essential to the task of tackling this challenge. Here, we investigate the dynamic contact patterns of the populations from anonymized, geo-localized mobility data and census and demographic data to create data-driven, agent-based contact networks. We then simulate the epidemic spread with a time-varying contagion model in ten large metropolitan counties in the United States and evaluate a combination of mobility reduction, mask use, and reopening policies. We find that our model captures the spatial-temporal and heterogeneous case trajectory within various counties based on dynamic population behaviors. Our results show that a decision-making tool that considers both economic cost and infection outcomes of policies can be informative in making decisions of local containment strategies for optimal balancing of economic slowdown and virus spread.

A network percolation-based contagion model of flood propagation and recession in urban road networks.

In this study, we propose a contagion model as a simple and powerful mathematical approach for predicting the spatial spread and temporal evolution of the onset and recession of floodwaters in urban road networks. A network of urban roads resilient to flooding events is essential for the provision of public services and for emergency response. The spread of floodwaters in urban networks is a complex spatial-temporal phenomenon. This study presents a mathematical contagion model to describe the spatial-temporal spread and recession process of floodwaters in urban road networks. The evolution of floods within networks can be captured based on three macroscopic characteristics-flood propagation rate ([Formula: see text]), flood incubation rate ([Formula: see text]), and recovery rate ([Formula: see text])-in a system of ordinary differential equations analogous to the Susceptible-Exposed-Infected-Recovered (SEIR) model. We integrated the flood contagion model with the network percolation process in which the probability of flooding of a road segment depends on the degree to which the nearby road segments are flooded. The application of the proposed model is verified using high-resolution historical data of road flooding in Harris County during Hurricane Harvey in 2017. The results show that the model can monitor and predict the fraction of flooded roads over time. Additionally, the proposed model can achieve 90% precision and recall for the spatial spread of the flooded roads at the majority of tested time intervals. The findings suggest that the proposed mathematical contagion model offers great potential to support emergency managers, public officials, citizens, first responders, and other decision-makers for flood forecast in road networks.

Abnormalities of autobiographical memory of patients with depressive disorders: a meta-analysis.

BACKGROUND: Previous studies on the autobiographical memory (AM) of depressed patients had inconsistent findings. Various severities of depression in patients in these studies may lead to conflicting results. However, the differences in the procedure of the autobiographical memory tests (AMTs) may also influence the AM results. OBJECTIVE: In this study, we analyse the results published so far to research the AM characteristics of patients with depressive disorders and identify moderators that affect the assessment results while using AMT in this field. METHOD: A systematic search was conducted using following databases: MEDLINE, PubMed, ScienceDirect, Cnki, and Google Scholar, yielding 22 studies of patients with depressive disorders and autobiographical memory published between 1986 and 2010. RESULTS: The results of meta-analysis showed that, compared with the control group, the patients with depressive disorders reported less specific AMs (g = -1.051) and more overgeneralized AMs (g = 1.115). The patients with depressive disorders also recalled more slowly (g = 0.400). The effect sizes of overgeneral memory could be predicted by the self-reported depression score of the depressed patients (B = -.329, p < .01). The mean effect sizes of AMT indices were affected by the AMT characteristics (i.e., number of cue word, max response time, prompting, presentation of cue word, taping, and so on). CONCLUSIONS: Our results suggest that overgeneralization and response lag are the AM deficits in patients with depressive disorders. The parameters of AMT are important factors, which are related to the inconsistency in the assessment of AM in patients with depressive disorders. Some recommendations on AMT and programme research design are given for future research.