Young Seog Yoon

Anomaly detection is a critical and challenging task that aims to identify data points deviating from normal patterns and distributions within a dataset. Various methods have been proposed using a one-class-one-model approach, but these techniques often face practical problems such as memory inefficiency and the requirement of sufficient data for training. In particular, few-shot anomaly detection presents significant challenges in industrial applications, where limited samples are available before mass production. In this paper, we propose a few-shot anomaly detection method that integrates adversarial training loss to obtain more robust and generalized feature representations. We utilize the adversarial loss previously employed in domain adaptation to align feature distributions between source and target domains, to enhance feature robustness and generalization in few-shot anomaly detection tasks. We hypothesize that adversarial loss is effective when applied to features that should have similar characteristics, such as those from the same layer in a Siamese network's parallel branches or input-output pairs of reconstruction-based methods. Experimental results demonstrate that the proposed method generally achieves better performance when utilizing the adversarial loss.

In this paper, we propose a novel MAC protocol, called DG-LoRa, for improving scalability in low power wide area networks. DG-LoRa is backward compatible with legacy LoRaWAN and adds new features, such as group acknowledgment transmissions in the time-synchronized frame structure that supports determinism on channel access. In DG-LoRa, the number of responses to data frames that are transmitted from end devices is maximized by allocating the spreading factor and timeslot in the frame structure. We evaluate the performance of DG-LoRa using the Monte-Carlo simulation and then compare it with the performance of legacy LoRaWAN in terms of data drop rate and the number of retransmissions. Our numerical results show that DG-LoRa supports approximately five times more connections to the LoRa network satisfying a 5% data drop rate. Also, it is observed that DG-LoRa enables low overhead by reducing the number of data frame retransmissions.