Deep Individual Active Learning: Safeguarding against Out-of-Distribution Challenges in Neural Networks.

Active learning (AL) is a paradigm focused on purposefully selecting training data to enhance a model's performance by minimizing the need for annotated samples. Typically, strategies assume that the training pool shares the same distribution as the test set, which is not always valid in privacy-sensitive applications where annotating user data is challenging. In this study, we operate within an individual setting and leverage an active learning criterion which selects data points for labeling based on minimizing the min-max regret on a small unlabeled test set sample. Our key contribution lies in the development of an efficient algorithm, addressing the challenging computational complexity associated with approximating this criterion for neural networks. Notably, our results show that, especially in the presence of out-of-distribution data, the proposed algorithm substantially reduces the required training set size by up to 15.4%, 11%, and 35.1% for CIFAR10, EMNIST, and MNIST datasets, respectively.

Low-Complexity Self-Interference Cancellation for Multiple Access Full Duplex Systems.

Self-interference occurs when there is electromagnetic coupling between the transmission and reception of the same node; thus, degrading the RX sensitivity to incoming signals. In this paper we present a low-complexity technique for self-interference cancellation in multiple carrier multiple access systems employing whole band direct to digital sampling. In this scenario, multiple users are simultaneously received and transmitted by the system at overlapping arbitrary bandwidths and powers. Traditional algorithms for self-interference mitigation based on recursive least squares (RLS) or least mean squares (LMS), fail to provide sufficient rejection, since the incoming signal is far from being spectrally flat, which is critical for their performance. The proposed algorithm mitigates the interference by modeling the incoming multiple user signal as an autoregressive (AR) process and jointly estimates the AR parameters and self-interference. The resulting algorithm can be implemented using a low-complexity architecture comprised of only two RLS modules. The novel algorithm further satisfies low latency constraints and is adaptive, supporting time varying channel conditions. We compare this to many self-interference cancellation algorithms, mostly adopted from the acoustic echo cancellation literature, and show significant performance gain.