In-memory computing uses resistive switching devices to perform computations within memory, eliminating the need for data transfer between memory and processing units. This approach addresses the "memory wall" limitation of traditional von Neumann architecture, where data movement is energy-intensive and time-consuming. Resistive switching devices, such as RRAM, PCM, MRAM, and FeRAM, offer area- and energy-efficient computation due to their two-terminal structure and direct data processing in memory. These devices enable digital, analog, and stochastic computing schemes, with RRAM being particularly promising for in-memory computing due to its fast switching speed, high density, and scalability. In-memory computing can be implemented through various schemes, including digital logic gates, analog computing using crosspoint arrays, and stochastic computing for random number generation. Digital computing with RRAM involves logic gates that perform operations based on resistive states, while analog computing uses crosspoint arrays to perform matrix-vector multiplication efficiently. Stochastic computing leverages the variability in resistive switching for random number generation, which is useful for cryptography and hardware security. Despite its potential, in-memory computing faces challenges such as switching variability, memory instability, and the need for efficient scaling. These issues affect the accuracy and reliability of computations, particularly in analog operations. To overcome these challenges, novel 3D array architectures and advanced materials are being explored to improve device performance and scalability. In-memory computing has the potential to revolutionize computing by reducing energy consumption and improving efficiency, making it a promising alternative to traditional von Neumann architectures. However, further research is needed to address interdisciplinary challenges in device optimization, circuit design, and system management to fully realize the benefits of in-memory computing.In-memory computing uses resistive switching devices to perform computations within memory, eliminating the need for data transfer between memory and processing units. This approach addresses the "memory wall" limitation of traditional von Neumann architecture, where data movement is energy-intensive and time-consuming. Resistive switching devices, such as RRAM, PCM, MRAM, and FeRAM, offer area- and energy-efficient computation due to their two-terminal structure and direct data processing in memory. These devices enable digital, analog, and stochastic computing schemes, with RRAM being particularly promising for in-memory computing due to its fast switching speed, high density, and scalability. In-memory computing can be implemented through various schemes, including digital logic gates, analog computing using crosspoint arrays, and stochastic computing for random number generation. Digital computing with RRAM involves logic gates that perform operations based on resistive states, while analog computing uses crosspoint arrays to perform matrix-vector multiplication efficiently. Stochastic computing leverages the variability in resistive switching for random number generation, which is useful for cryptography and hardware security. Despite its potential, in-memory computing faces challenges such as switching variability, memory instability, and the need for efficient scaling. These issues affect the accuracy and reliability of computations, particularly in analog operations. To overcome these challenges, novel 3D array architectures and advanced materials are being explored to improve device performance and scalability. In-memory computing has the potential to revolutionize computing by reducing energy consumption and improving efficiency, making it a promising alternative to traditional von Neumann architectures. However, further research is needed to address interdisciplinary challenges in device optimization, circuit design, and system management to fully realize the benefits of in-memory computing.