Convolutional neural networks (CNNs) have been widely used in image recognition and processing tasks. Memristor-based CNNs accumulate the advantages of emerging memristive devices, such as nanometer ...critical dimensions, low power consumption, and functional similarity to biological synapses. Most studies on memristor-based CNNs use either software models of memristors for simulation analysis or full hardware CNN realization. Here, we propose a hybrid CNN, consisting of a hardware fixed pre-trained and explainable feature extractor and a trainable software classifier. The hardware part was realized on passive crossbar arrays of memristors based on nanocomposite (Co-Fe-B)x(LiNbO3)100−x structures. The constructed 2-kernel CNN was able to classify the binarized Fashion-MNIST dataset with ~ 84% accuracy. The performance of the hybrid CNN is comparable to the other reported memristor-based systems, while the number of trainable parameters for the hybrid CNN is substantially lower. Moreover, the hybrid CNN is robust to the variations in the memristive characteristics: dispersion of 20% leads to only a 3% accuracy decrease. The obtained results pave the way for the efficient and reliable realization of neural networks based on partially unreliable analog elements.
Convolutional neural networks (CNNs) have been widely used in image recognition and processing tasks. Memristor-based CNNs accumulate the advantages of emerging memristive devices, such as nanometer ...critical dimensions, low power consumption, and functional similarity to biological synapses. Most studies on memristor-based CNNs use either software models of memristors for simulation analysis or full hardware CNN realization. Here, we propose a hybrid CNN, consisting of a hardware fixed pre-trained and explainable feature extractor and a trainable software classifier. The hardware part was realized on passive crossbar arrays of memristors based on nanocomposite (Co-Fe-B)sub.x(LiNbOsub.3)sub.100−x structures. The constructed 2-kernel CNN was able to classify the binarized Fashion-MNIST dataset with ~ 84% accuracy. The performance of the hybrid CNN is comparable to the other reported memristor-based systems, while the number of trainable parameters for the hybrid CNN is substantially lower. Moreover, the hybrid CNN is robust to the variations in the memristive characteristics: dispersion of 20% leads to only a 3% accuracy decrease. The obtained results pave the way for the efficient and reliable realization of neural networks based on partially unreliable analog elements.
The paper discusses a new method for diagnosing atherosclerosis using a pneumatic arterial blood pressure sensor previously developed by the authors hereof. The possibility of applying a pneumatic ...sensor to measure the pulse wave transit time referring to a synchronous ECG is treated herein. The specification of the method consisting in the selection of the characteristic moment of the pulse wave as the timestamp, when measuring the signal transit time in relation to the R-peak of the synchronous ECG, is justified hereby. The averaged values of the wave transit time at different points of the artery, taking into account the variability of the front delay values, are used to directly determine the pulse wave propagation velocity in the area between the measurement points.
Low-power X-ray tubes: the current status Bugaev, Aleksandr S.; Eroshkin, Pavel A.; Romanko, Vasilij A. ...
Uspehi fiziceskih nauk,
2013, Volume:
183, Issue:
7
Journal Article
PDF
