No abstract

Memristor is a fundamental circuit element in addition to resistor, capacitor, and inductor. As it can remember its resistance state even encountering a power off, memristor has recently received widespread applications from non-volatile memory to neural networks. The current memristor family mainly comprises resistive memristor, polymeric memristor, ferroelectric memristor, manganite memristor, resonant-tunneling diode memristor, and spintronic memristor in terms of the materials the device is made of. In order to help researcher better understand the physical principles of the memristor, and thus to provide a promising prospect for memristor devices, this paper presents an overview of memristor materials properties, switching mechanisms, and potential applications. The performance comparison among different memristor members is also given.

Today the influence of non-volatile solid-state memories on persons' lives has become more prominent because of their non-volatility, low data latency, and high robustness. As a pioneering technology that is representative of non-volatile solidstate memories, flash memory has recently seen widespread application in many areas ranging from electronic appliances, such as cell phones and digital cameras, to external storage devices such as universal serial bus (USB) memory. Moreover, owing to its large storage capacity, it is expected that in the near future, flash memory will replace hard-disk drives as a dominant technology in the mass storage market, especially because of recently emerging solid-state drives. However, the rapid growth of the global digital data has led to the need for flash memories to have larger storage capacity, thus requiring a further downscaling of the cell size. Such a miniaturization is expected to be extremely difficult because of the well-known scaling limit of flash memories. It is therefore necessary to either explore innovative technologies that can extend the areal density of flash memories beyond the scaling limits, or to vigorously develop alternative non-volatile solid-state memories including ferroelectric random-access memory, magnetoresistive random-access memory, phase-change random-access memory, and resistive random-access memory. In this paper, we review the physical principles of flash memories and their technical challenges that affect our ability to enhance the storage capacity. We then present a detailed discussion of novel technologies that can extend the storage density of flash memories beyond the commonly accepted limits. In each case, we subsequently discuss the physical principles of these new types of non-volatile solid-state memories as well as their respective merits and weakness when utilized for data storage applications. Finally, we predict the future prospects for the aforementioned solid-state memories for the next generation of data-storage devices based on a comparison of their performance.

Phase-change probe memory using Ge 2 Sb 2 Te 5 has been considered as one of the promising candidates as next-generation data storage device due to its ultra-high density, low energy consumption, short access time and long retention time. In order to utmostly mimic the practical setup, and thus fully explore the potential of phase-change probe memory for 10 Tbit/in 2 target, some advanced modeling techniques that include threshold-switching, electrical contact resistance, thermal boundary resistance and crystal nucleation-growth, are introduced into the already-established electrothermal model to simulate the write and read performance of phasechange probe memory using an optimal media stack design. The resulting predictions clearly demonstrate the capability of phase-change probe memory to record 10 Tbit/in 2 density under pico Joule energy within micro second period.

Background: Medical image segmentation plays a vital role in computer-aided diagnosis (CAD) systems. Both convolutional neural networks (CNNs) with strong local information extraction capacities and transformers with excellent global representation capacities have achieved remarkable performance in medical image segmentation. However, because of the semantic differences between local and global features, how to combine convolution and transformers effectively is an important challenge in medical image segmentation. Methods: In this paper, we proposed TransConver, a U-shaped segmentation network based on convolution and transformer for automatic and accurate brain tumor segmentation in MRI images. Unlike the recently proposed transformer and convolution based models, we proposed a parallel module named transformerconvolution inception (TC-inception), which extracts local and global information via convolution blocks and transformer blocks, respectively, and integrates them by a cross-attention fusion with global and local feature (CAFGL) mechanism. Meanwhile, the improved skip connection structure named skip connection with cross-attention fusion (SCCAF) mechanism can alleviate the semantic differences between encoder features and decoder features for better feature fusion. In addition, we designed 2D-TransConver and 3D-TransConver for 2D and 3D brain tumor segmentation tasks, respectively, and verified the performance and advantage of our model through brain tumor datasets. Results: We trained our model on 335 cases from the training dataset of MICCAI BraTS2019 and evaluated the model's performance based on 66 cases from MICCAI BraTS2018 and 125 cases from MICCAI BraTS2019. Our TransConver achieved the best average Dice score of 83.72% and 86.32% on BraTS2019 and BraTS2018, respectively. Conclusions: We proposed a transformer and convolution parallel network named TransConver for brain tumor segmentation. The TC-Inception module effectively extracts global information while retaining local details. The experimental results demonstrated that good segmentation requires the model to extract local fine-grained details and global semantic information simultaneously, and our TransConver effectively improves the accuracy of brain tumor segmentation.