arXiv subjects

Yizhi Zhao

Publications and source records attributed to Yizhi Zhao.

5 recordsLinked to original sources

arXiv2023

Semi-supervised Cell Recognition under Point Supervision

Cell recognition is a fundamental task in digital histopathology image analysis. Point-based cell recognition (PCR) methods normally require a vast number of annotations, which is extremely costly, time-consuming and labor-intensive. Semi-supervised learning (SSL) can provide a shortcut to make full use of cell information in gigapixel whole slide images without exhaustive labeling. However, research into semi-supervised point-based cell recognition (SSPCR) remains largely overlooked. Previous SSPCR works are all built on density map-based PCR models, which suffer from unsatisfactory accuracy, slow inference speed and high sensitivity to hyper-parameters. To address these issues, end-to-end PCR models are proposed recently. In this paper, we develop a SSPCR framework suitable for the end-to-end PCR models for the first time. Overall, we use the current models to generate pseudo labels for unlabeled images, which are in turn utilized to supervise the models training. Besides, we introduce a co-teaching strategy to overcome the confirmation bias problem that generally exists in self-training. A distribution alignment technique is also incorporated to produce high-quality, unbiased pseudo labels for unlabeled data. Experimental results on four histopathology datasets concerning different types of staining styles show the effectiveness and versatility of the proposed framework. Code is available at \textcolor{magenta}{\url{https://github.com/windygooo/SSPCR}

cs.CV↗

arXiv2021

A Proof of Non-stationary Channel Polarization

In this letter, we consider the proof of non-stationary channel polarization theory. First we construct a multi-channel stochastic process for the non-stationary channel polarization operation. Then based on this stochastic process, we extend Arıkan's standard martingale proof method on the average channel capacity and average channel Bhattacharyya parameter, by which we have proved the non-stationary channel polarization theory.

cs.IT↗

arXiv2020

Secure Polar Coding for Adversarial Wiretap Channel

The adversarial wiretap channel (AWTC) model is a secure communication model in which adversary can directly read and write the transmitted bits in legitimate communication with fixed fractions. In this paper we propose a secure polar coding scheme to provide secure and reliable communication over the AWTC model. For the adversarial reading and writing action, we present a $ρ$ equivalent channel block and study its transformation under the channel polarization operation. We find that the generated channels are polarized in the sense that part of them are full-noise channels with probability almost $1$ and the rest part of them are noiseless channels with probability almost $1$. Based on this result, we polarize both equivalent channel blocks of adversarial reading and writing, and then construct a secure polar coding scheme by applying the multi-block chaining structure on the polarized equivalent blocks. Theoretically we prove that when block length $N$ goes infinity, the proposed scheme achieves the secrecy capacity of the AWTC model under both reliability and strong security criterions. Then by simulations, we prove that the proposed scheme can provide secure and reliable communication over AWTC model.

cs.IT↗

arXiv2020

CSI Learning Based Active Secure Coding Scheme For Detectable Wiretap Channel

In this paper, we consider the problem of secure and reliable communication with uncertain channel state information (CSI) and present a new solution named active secure coding which combines the machine learning methods with the traditional physical layer secure coding scheme. First, we build a detectable wiretap channel model by combining the hidden Markov model with the compound wiretap channel model, in which the varying of channel block CSI is a Markov process and the detected information is a stochastic emission from the current CSI. Next, we present a CSI learning scheme to learn the CSI from the detected information by the Baum-Welch and Viterbi algorithms. Then we construct explicit secure polar codes based on the learned CSI, and combine it with the CSI learning scheme to form the active secure polar coding scheme. Simulation results show that an acceptable level of reliability and security can be achieved by the proposed active secure polar coding scheme.

cs.IT↗

arXiv2020

Secure Polar Coding with Delayed Wiretapping Information

In this paper, we investigate the secure coding issue for a wiretap channel model with fixed main channel and varying wiretap channel, by assuming that legitimate parties can obtain the wiretapping channel state information (CSI) after some time delay. For the symmetric degraded delay CSI case, we present an explicit weak security scheme by constructing secure polar codes on a one-time pad chaining structure, and prove its weak security, reliability and capability of approaching the secrecy capacity of perfect CSI case with delay CSI assumption. Further for the symmetric no-degraded delay CSI case, we present a modified multi-block chaining structure in which the original subset of frozen bit is designed for conveying functional random bits securely. Then we combine this modified multi-block chaining structure with the weak security scheme to construct an explicit strong security polar coding scheme, and prove its strong security, reliability and also the capability of approaching the secrecy capacity of perfect CSI case with delay CSI assumption. At last, we carry out stimulations to prove the performance of both secure schemes.

cs.IT↗