This paper investigates the issue of uplink integrated sensing and communication (ISAC) in 6G wireless networks, where the sensing echo signal and the communication signal are received simultaneously at the base station (BS). To mitigate mutual interference between sensing and communication caused by shared spectrum and hardware resources, the authors propose a joint sensing transmit waveform and communication receive beamforming design aimed at maximizing the weighted sum of normalized sensing rate and normalized communication rate. This is formulated as a computationally complex non-convex optimization problem, which is challenging to solve with conventional methods. To address this, the authors develop a deep learning (DL)-based scheme to enhance ISAC performance. Both theoretical analysis and simulation results confirm the effectiveness and robustness of the proposed DL-based scheme for ISAC in 6G wireless networks. The paper introduces a general design framework for non-orthogonal uplink ISAC systems, where a dual-function BS is deployed to sense nearby targets and serve multiple communication users (CUs) simultaneously. The authors propose a joint sensing transmit waveform and communication receive beamforming design to mitigate mutual interference for simultaneous sensing information extraction and communication signal decoding. This is formulated as a weighted sum of normalized sensing rate and normalized communication rate maximization problem. To reduce design complexity, equivalent problem transformations are made, and a customized deep neural network (DNN) called "ISACNN" is designed with unsupervised learning based on the characteristics of non-orthogonal uplink ISAC systems. The paper presents a DL-based scheme for uplink ISAC, which is designed to reduce mutual interference and achieve desired ISAC performance. The proposed DL-based scheme is validated through numerical simulations, showing its effectiveness in improving the performance of ISAC systems. The contributions of the paper include the development of a general design framework for non-orthogonal uplink ISAC systems, the proposal of a joint sensing transmit waveform and communication receive beamforming design, and the design of a customized deep neural network for ISAC systems. The paper also discusses the computational complexity of the proposed DL-based scheme and presents simulation results demonstrating its effectiveness in improving the performance of ISAC systems.This paper investigates the issue of uplink integrated sensing and communication (ISAC) in 6G wireless networks, where the sensing echo signal and the communication signal are received simultaneously at the base station (BS). To mitigate mutual interference between sensing and communication caused by shared spectrum and hardware resources, the authors propose a joint sensing transmit waveform and communication receive beamforming design aimed at maximizing the weighted sum of normalized sensing rate and normalized communication rate. This is formulated as a computationally complex non-convex optimization problem, which is challenging to solve with conventional methods. To address this, the authors develop a deep learning (DL)-based scheme to enhance ISAC performance. Both theoretical analysis and simulation results confirm the effectiveness and robustness of the proposed DL-based scheme for ISAC in 6G wireless networks. The paper introduces a general design framework for non-orthogonal uplink ISAC systems, where a dual-function BS is deployed to sense nearby targets and serve multiple communication users (CUs) simultaneously. The authors propose a joint sensing transmit waveform and communication receive beamforming design to mitigate mutual interference for simultaneous sensing information extraction and communication signal decoding. This is formulated as a weighted sum of normalized sensing rate and normalized communication rate maximization problem. To reduce design complexity, equivalent problem transformations are made, and a customized deep neural network (DNN) called "ISACNN" is designed with unsupervised learning based on the characteristics of non-orthogonal uplink ISAC systems. The paper presents a DL-based scheme for uplink ISAC, which is designed to reduce mutual interference and achieve desired ISAC performance. The proposed DL-based scheme is validated through numerical simulations, showing its effectiveness in improving the performance of ISAC systems. The contributions of the paper include the development of a general design framework for non-orthogonal uplink ISAC systems, the proposal of a joint sensing transmit waveform and communication receive beamforming design, and the design of a customized deep neural network for ISAC systems. The paper also discusses the computational complexity of the proposed DL-based scheme and presents simulation results demonstrating its effectiveness in improving the performance of ISAC systems.