Efficient Decomposition of Unitary Matrices in Quantum Circuit Compilers

Unitary decomposition is a widely used method to map quantum algorithms to an arbitrary set of quantum gates. Efficient implementation of this decomposition allows for the translation of bigger unitary gates into elementary quantum operations, which is key to executing these algorithms on existing quantum computers. The decomposition can be used as an aggressive optimization method for the whole circuit, as well as to test part of an algorithm on a quantum accelerator. For the selection and implementation of the decomposition algorithm, perfect qubits are assumed. We base our decomposition technique on Quantum Shannon Decomposition, which generates O(344n) controlled-not gates for an n-qubit input gate. In addition, we implement optimizations to take advantage of the potential underlying structure in the input or intermediate matrices, as well as to minimize the execution time of the decomposition. Comparing our implementation to Qubiter and the UniversalQCompiler (UQC), we show that our implementation generates circuits that are much shorter than those of Qubiter and not much longer than the UQC. At the same time, it is also up to 10 times as fast as Qubiter and about 500 times as fast as the UQC.

A novel long short-term memory network for lithium-ion battery health diagnosis using charging curve

Lithium-ion batteries are widely used as the power source in electric vehicles. The state of health (SOH) diagnosis is very important for the safety and storage capacity of lithium-ion batteries. In order to accurately and robustly estimate lithium-ion battery SOH, a novel long short-term memory network (LSTM) based on the charging curve is proposed for SOH estimation in this work. Firstly, aging features that reflect the battery degradation phenomenon are extracted from the charging curves. Then, considering capture the long-term tendency of battery degradation, some improvements are made in the proposed LSTM model. The connection between the input gate and the output gate is added to better control output information of the memory cell. Meanwhile, the forget gate and input gate are coupled into a single update gate for selectively forgetting before the accumulation of information. To achieve more reliability and robustness of the SOH estimation method, the improved LSTM network is adaptively trained online by using a particle filter. Furthermore, to verify the effectiveness of the proposed method, we compare the proposed method with two data-driven methods on the public battery data set and the commercial battery data set. Experimental results demonstrate the proposed method can obtain the highest SOH accuracy.

A new image encryption algorithm based on the OF-LSTMS and chaotic sequences

In this paper, a novel image encryption algorithm based on the Once Forward Long Short Term Memory Structure (OF-LSTMS) and the Two-Dimensional Coupled Map Lattice (2DCML) fractional-order chaotic system is proposed. The original image is divided into several image blocks, each of which is input into the OF-LSTMS as a pixel sub-sequence. According to the chaotic sequences generated by the 2DCML fractional-order chaotic system, the parameters of the input gate, output gate and memory unit of the OF-LSTMS are initialized, and the pixel positions are changed at the same time of changing the pixel values, achieving the synchronization of permutation and diffusion operations, which greatly improves the efficiency of image encryption and reduces the time consumption. In addition the 2DCML fractional-order chaotic system has better chaotic ergodicity and the values of chaotic sequences are larger than the traditional chaotic system. Therefore, it is very suitable to image encryption. Many simulation results show that the proposed scheme has higher security and efficiency comparing with previous schemes.

The role of the dentate gyrus in mnemonic functions

The hippocampus is decisive for the storage of conscious memories. Current theories suggest that experience-dependent modifications in excitation–inhibition balance enable a select group of neurons to form a new cell association during learning which represents the new memory trace. It was further proposed that particularly GABAergic-inhibitory interneurons have a large impact on population activity in neuronal networks by means of their inhibitory output synapses. They synchronize active principal cells at high frequencies, thereby supporting their binding to cell assemblies to jointly encode information. However, how cell associations emerge in space and time and how interneurons may contribute to this process is still largely unknown. We started to address this fundamental question in the dentate gyrus (DG) as the input gate of the hippocampus, which has an indispensable role in conscious memory formation. We used a combination of in vivo chronic two-photon imaging of population activity in the DG and the hippocampal areas CA1–3 of mice exposed to a virtual reality, in which they perform a goal-oriented spatial memory tasks, with high-density in vivo recordings and multiple whole-cell recordings in acute slice preparations, to determine how memory engrams emerge during learning. We further examine how GABAergic interneurons may contribute to this process. We believe that these lines of research will add to a better understanding on the mechanisms of memory formation in cortical networks.

Implementasi Deep Learning Menggunakan Arsitektur Long Short Term Memory(LSTM) Untuk Prediksi Curah Hujan Kota Malang

AbstrakTidak selamanya cuaca di Indonesia berjalan dengan normal atau sesuai dengan musimnya, cuaca sering berubah secara tiba-tiba setiap saat karena ada faktor-faktor yang mempengaruhi penurunan dan peningkatan curah hujan. perkiraan cuaca sangatlah dibutuhkan dan sangat bermanfaat olah berbagai pihak karena bisa menjadi acuan bagi berbagai kalangan untuk menjalani kegiatan mereka sehari-hari. Penelitian dilakukan menggunakan metode Deep Learning karena dari beberapa penelitian sebelumnya yang menggunakan Deep Learning dalam kasus yang berbeda mampu menghasilkan akurasi diatas 85%. Deep learning adalah jaringan yang terdiri dari beberapa layer. Layer-layer tersebut berasal dari kumpulan node-node. Arsitektur yang digunakan yaitu Long Short Term Memory(LSTM) karena pada penelitian-penelitian sebelumnya menggunakan LSTM dalam kasus yang berbeda mendapat hasil yang baik yaitu RME yang dihasilkan kecil. LSTM memiliki struktur seperti rantai dan struktur pada tiap sel terdapat 3 gate yaitu forget gate, input gate, dan output gate. Oleh karena itu, perhitungan yang dilakukan lebih kompleks ditambah lagi dengan Deep Learning diharapkan mendapat hasil yang lebih akurat. Data yang digunakan yaitu data curah hujan kota Malang yang berasal dari BMKG. Abstract The weather in Indonesia does not always run normally or in accordance with the season, the weather often changes suddenly at any time because there are factors that affect the decrease and increase in rainfall. weather forecasts are needed and very useful if the various parties because it can be a reference for various circles to undergo their daily activities. The study was conducted using Deep Learning method because of some previous research using Deep Learning in different cases able to produce accuracy above 85%. Deep learning is a network consisting of several layers. The layers are derived from a collection of nodes. The architecture used is Long Short Term Memory (LSTM) because in previous studies using LSTM in different case got good result that is small generated RME. LSTM has a structure like chains and structures in each cell there are 3 gates of forget gate, input gate, and output gate. Therefore, the calculations performed more complex plus the Deep Learning is expected to get more accurate results. The data used is the rainfall data of Malang city that comes from BMKG.

A Novel Peripheral Circuit for SWSFET Based Multivalued Static Random-Access Memory

This paper presents the peripheral circuitry for a multivalued static random-access memory (SRAM) based on 2-bit CMOS cross-coupled inverters using spatial wavefunction switched (SWS) field effect transistors (SWSFETs). The novel feature is a two quantum well/quantum dot channel n-SWSFET access transistor. The reduction in area with four-bit storage-per-cell increases the memory density and efficiency of the SRAM array. The SWSFET has vertically stacked two-quantum well/quantum dot channels between the source and drain regions. The upper or lower quantum charge locations in the channel region is based on the input gate voltage. The analog behavioral modeling (ABM) of the SWSFET device is done using conventional BSIM 3V3 device parameters in 90 nm technology. The Cadence circuit simulations for the proposed memory cell and addressing/peripheral circuitry are presented.

A Novel Addressing Circuit for SWS-FET Based Multivalued Dynamic Random-Access Memory Array

Multivalued memory increases the bits-per-cell storage capacity over conventional one transistor (1T) MOS based dynamic random-access memory (DRAM) by storing more than two data signal levels in each unit memory cell. A spatial wavefunction switched (SWS) field effect transistor (FET) has two vertically stacked quantum-well/quantum-dot channels between the source and drain regions. The charge location in upper or lower quantum channel region is based on the input gate voltage. A multivalued DRAM that can store more than two bits-per-cell was implemented by using one SWS-FET (1T) device and two capacitors (2C) connected to each source regions of the SWS-FET device. This paper proposes the architecture and design of peripheral circuitry that includes row/column address decoding and sensing circuit for a multivalued DRAM crossbar arrays. The SWS-FET device was modeled using analog behavioral modeling (ABM) with two transistors using conventional BSIM 3V3 device parameters in 90 nm technology. The Cadence circuit schematic simulations are presented. A compact multivalued DRAM architecture presents a new paradigm in terms of application in Neural systems that demand storage of multiple valued levels.

Reading selectively via Binary Input Gated Recurrent Unit

Recurrent Neural Networks (RNNs) have shown great promise in sequence modeling tasks. Gated Recurrent Unit (GRU) is one of the most used recurrent structures, which makes a good trade-off between performance and time spent. However, its practical implementation based on soft gates only partially achieves the goal to control information flow. We can hardly explain what the network has learnt internally. Inspired by human reading, we introduce binary input gated recurrent unit (BIGRU), a GRU based model using a binary input gate instead of the reset gate in GRU. By doing so, our model can read selectively during interference. In our experiments, we show that BIGRU mainly ignores the conjunctions, adverbs and articles that do not make a big difference to the document understanding, which is meaningful for us to further understand how the network works. In addition, due to reduced interference from redundant information, our model achieves better performances than baseline GRU in all the testing tasks.

A Deep Learning Method for Bearing Fault Diagnosis through Stacked Residual Dilated Convolutions

Real-time monitoring and fault diagnosis of bearings are of great significance to improve production safety, prevent major accidents, and reduce production costs. However, there are three primary concerns in the current research, namely real-time performance, effectiveness, and generalization performance. In this paper, a deep learning method based on stacked residual dilated convolutional neural network (SRDCNN) is proposed for real-time bearing fault diagnosis, which is subtly combined by the dilated convolution, the input gate structure of long short-term memory network (LSTM) and the residual network. In the SRDCNN model, the dilated convolution is used to exponentially increase the receptive field of convolution kernel and extract features from the sample with more points, alleviating the influence of randomness. The input gate structure of LSTM could effectively remove noise and control the entry of information contained in the input sample. Meanwhile, the residual network is introduced to overcome the problem of vanishing gradients caused by the deeper structure of the neural network, hence improving the overall classification accuracy. The experimental results indicate that compared with three excellent models, the proposed SRDCNN model has higher denoising ability and better workload adaptability.

Image Matching via Loopy RNN

Most existing matching algorithms are one-off algorithms, i.e., they usually measure the distance between the two image feature representation vectors for only one time. In contrast, human's vision system achieves this task, i.e., image matching, by recursively looking at specific/related parts of both images and then making the final judgement. Towards this end, we propose a novel loopy recurrent neural network (Loopy RNN), which is capable of aggregating relationship information of two input images in a progressive/iterative manner and outputting the consolidated matching score in the final iteration. A Loopy RNN features two uniqueness. First, built on conventional long short-term memory (LSTM) nodes, it links the output gate of the tail node to the input gate of the head node, thus it brings up symmetry property required for matching. Second, a monotonous loss designed for the proposed network guarantees increasing confidence during the recursive matching process. Extensive experiments on several image matching benchmarks demonstrate the great potential of the proposed method.