While machine learning algorithms hold promise for personalised medicine, their clinical adoption remains limited, partly due to biases that can compromise the reliability of predictions. In this article, we focus on sample selection bias (SSB), a specific type of bias where the study population is less representative of the target population, leading to biased and potentially harmful decisions. Despite being well-known in the literature, SSB remains scarcely studied in machine learning for healthcare. Moreover, the existing machine learning techniques try to correct the bias mostly by balancing distributions between the study and the target populations, which may result in a loss of predictive performance. To address these problems, our study illustrates the potential risks associated with SSB by examining SSB’s impact on the performance of machine learning algorithms. Most importantly, we propose a new research direction for addressing SSB, based on the target population identification rather than the bias correction. Specifically, we propose two independent networks (T-Net) and a multitasking network (MT-Net) for addressing SSB, where one network/task identifies the target subpopulation which is representative of the study population and the second makes predictions for the identified subpopulation. Our empirical results with synthetic and semi-synthetic datasets highlight that SSB can lead to a large drop in the performance of an algorithm for the target population as compared with the study population, as well as a substantial difference in the performance for the target subpopulations that are representative of the selected and the non-selected patients from the study population. Furthermore, our proposed techniques demonstrate robustness across various settings, including different dataset sizes, event rates and selection rates, outperforming the existing bias correction techniques.

This paper presents a stochastic geometry model for studying spatiotemporal base station cooperation in a cellular network. Unlike most previous related works in the literature, the paper focuses on the coverage probability at the typical cellcorner user, i.e., a user that is equidistant to its three neighboring base stations (BSs). Three schemes are considered: joint transmission, base station silencing, and the Alamouti space-time code. In addition, BSs are allowed to cooperatively retransmit data across consecutive time slots, such that the cell-corner user can benefit from time diversity. For all these schemes, an expression for the coverage probability at the cell-corner user is derived. The best performing scheme is shown to be the Alamouti code, when only two BSs cooperate. Otherwise, cooperative retransmission performs the best in the high-coverage regime, while joint transmission without retransmission outperforms the remaining schemes in all other regimes. These results extend previous findings on spatiotemporal base station cooperation that only applied to the typical user in a cellular network.

This paper studies downlink communication in a heterogeneous cellular network where a set of geographically separated base stations (BSs) cooperates in transmitting data to a common receiver. If a decoding error occurs, data is cooperatively retransmitted by a possibly different set of BSs, such that the receiver can benefit from spatiotemporal BS cooperation. Specific cooperation techniques studied in this paper include joint transmission, base station silencing, and the Alamouti space-time code. Using tools from stochastic geometry, the coverage probability at the typical user is characterized as an integral function of the network parameters and the sets of cooperating BSs. The expressions derived reveal the existence of two qualitatively different operating regimes. In the high-coverage regime, the typical user is diversity-limited, so cooperation techniques exploiting spatiotemporal diversity are highly effective in increasing coverage. It is shown that retransmissions always yield time diversity, while channel state information at the transmitters is required to harvest spatial diversity via joint transmission. In the low-coverage regime, on the other hand, the typical user is interference-limited, so cooperation techniques such as joint transmission and base station silencing are effective in increasing coverage as they suppress part of the interference power.

This paper studies spatiotemporal cooperation between base stations in the presence of interference in a heterogeneous cellular network. The focus of the paper is the cooperative retransmission scenario, where a set of randomly located base stations that are selected based on their average received powers, possibly belonging to different network tiers, jointly transmit data in each transmission. If a decoding error occurs in the first transmission, data is cooperatively retransmitted by a possibly different set of BSs, such that the receiver can benefit from spatiotemporal BS cooperation. Using tools from stochastic geometry, the coverage probability at the typical user is characterized as an integral function of the network parameters and the sets of cooperating BSs. An integral expression for the coverage probability is also derived for the case when the typical user is able to perform maximum ratio combining (MRC) of the received copies in two transmissions. Numerical evaluation shows that in the high coverage regime spatiotemporal cooperation outperforms joint transmission with no retransmissions and that the gains extend to all regimes if the user has MRC capability.

Motivated by the ongoing discussion on coordinated multipoint in wireless cellular standard bodies, this paper considers the problem of base station cooperation in the downlink of heterogeneous cellular networks. The focus of this paper is the joint transmission scenario, where an ideal backhaul network allows a set of randomly located base stations, possibly belonging to different network tiers, to jointly transmit data, to mitigate intercell interference and hence improve coverage and spectral efficiency. Using tools from stochastic geometry, an integral expression for the network coverage probability is derived in the scenario where the typical user located at an arbitrary location, i.e., the general user, receives data from a pool of base stations that are selected based on their average received power levels. An expression for the coverage probability is also derived for the typical user located at the point equidistant from three base stations, which we refer to as the worst case user. In the special case where cooperation is limited to two base stations, numerical evaluations illustrate absolute gains in coverage probability of up to 17% for the general user and 24% for the worst case user compared with the noncooperative case. It is also shown that no diversity gain is achieved using noncoherent joint transmission, whereas full diversity gain can be achieved at the receiver if the transmitting base stations have channel state information.