ReRAM crossbar-based computing-in-memory (CIM) systems offer computational efficiency but suffer from significant accuracy degradation under stuck-at faults (SAFs). Conventional approaches like retraining-based methods fail to effectively generalize across diverse SAFs ratios. To address this challenge, we propose the Mask-based Meta Learning (MML) framework, leveraging meta-learning’s multi-task generalization capability to achieve robust performance of varying SAFs scenarios. Within the MML framework, a SAFs mask-based task formulation is used to create meta-learning tasks based on SAFs masks rather than dataset. Second, we define a new meta-learning objective by integrating different SAFs masks into the meta loss. Finally, we develop a SAFs-sensitivity guided weight importance search algorithm and dynamically expands the adjustment range of crucial weights using the ReRAM array’s redundant cells to further enhance model performance. Experimental results show that our method demonstrates superior robustness and generalization performance over the state-of-the-art robust training approaches. Moreover, our method can maintain high accuracy across various SAFs ratios while the accuracy of other approaches illustrates large fluctuations.

The rise of data-intensive AI workloads has exacerbated the "memory wall" bottleneck. Digital Compute-in-Memory (DCiM) using SRAM offers a scalable solution, but its vast design space makes manual design impractical, creating a need for automated compilers. A key opportunity lies in approximate computing, which leverages the error tolerance of AI applications for significant energy savings. However, existing DCiM compilers focus on exact arithmetic, failing to exploit this optimization. This paper introduces OpenACM, the first open-source, accuracy-aware compiler for SRAM-based approximate DCiM architectures. OpenACM bridges the gap between application error tolerance and hardware automation. Its key contribution is an integrated library of accuracy-configurable multipliers (exact, tunable approximate, and logarithmic), enabling designers to make fine-grained accuracy-energy trade-offs. The compiler automates the generation of the DCiM architecture, integrating a transistor-level customizable SRAM macro with variation-aware characterization into a complete, open-source physical design flow based on OpenROAD and the FreePDK45 library. This ensures full reproducibility and accessibility, removing dependencies on proprietary tools. Experimental results on representative convolutional neural networks (CNNs) demonstrate that OpenACM achieves energy savings of up to 64% with negligible loss in application accuracy. The framework is available on OpenACM:URL.

The design of MEMS sensor presents a significant challenge in identifying feasible structures that align with specific performance criteria. Traditionally, this process demands extensive design expertise and iterative simulations, leading to time-intensive workflows. While recent advancements have introduced deep learning (DL) models to expedite this process, they are limited to handling simple scenarios with precise performance values and fixed dimensions as inputs, often overlooking the uncertainty inherent in real design scenarios, such as vague range requirements and variable input dimensions. To address this issue, this study introduces a novel DL-based design model along with corresponding modeling strategies. The proposed model consists of a search network (SN), a validation network (VN), and a precision optimizer (PO). Initially, design requirements of various types and dimensions are transformed into a standardized input vector to address diverse design scenarios, which is then processed by the SN to generate a feasible structure. The VN, trained prior to the SN, validates the structure and generates training data for the SN. In cases where the model output fails to sufficiently align with the requirements, the PO is deployed to minimize the design error. Validation of the proposed model was conducted using a piezoresistive acceleration sensor across 100000 distinct design requirements. The results demonstrate an overall design accuracy (DA) of 92.64% on the testing data. Following 1000 iterations leveraging the proposed PO, the DA improves to 93.84%. Notably, each design iteration and optimization using the PO only requires approximately 0.1 ms, significantly boosting the design efficiency of MEMS sensors.

To achieve the desired characteristics for MEMS sensors, the traditional design process obtains the geometrical parameters based on complex theoretical calculations and interactive finite-element (FE) simulations, which are time consuming and data consuming. To solve the above problems, a data-driven bidirectional design approach based on the deep learning (DL) method is introduced to improve the design efficiency of MEMS sensors in this work. By using the piezoresistive acceleration sensor as a design example, the forward artificial neural network (ANN) with the sensor geometrical parameters as the input and the sensor performance as the output is trained and realized by using 1000 groups of data collected through FE simulation. This forward ANN can accurately predict the sensor performance, including the measurement range, sensitivity, and resonant frequency. In addition, the inverse ANN with the sensor performance as the input and the sensor geometrical parameters as the output is also achieved by using a tandem network. This inverse ANN can provide the geometrical parameters directly and instantly according to the target performance. Both the forward and inverse networks cost only about 6 ms for each task and the mean relative errors are less than 3%. The high efficiency and low relative error indicate that DL is a promising approach to improve the design efficiency for MEMS sensors.

Primary-side regulation (PSR) flyback topology with discontinuous conduction modulation (DCM) mode has been widely used in low-power application due to its high stability, wide input voltage range, low cost, and low standby power. To improve the output power, continuous conduction modulation (CCM) mode is always introduced. However, in CCM mode, conventional PSR voltage-sampling method cannot realize high precision of output voltage or the voltage sampling method requires complex control of high cost. In this paper, a simple digital voltage sampling scheme with two reference voltages outputted from a digital-to-analog converter (DAC) is put forward to trace the “accurate-point” voltage for DCM and CCM modes. In CCM mode, the sampling voltage error of the “knee-point” voltage can be compensated just by inserting DCM switching cycle to obtain the “accurate-point” voltage. Voltage difference between DCM mode and CCM mode will be eliminated, and high precision is easily realized. The proposed method is verified in a 20V, 65W PSR flyback converter. The sampling method only requires a low speed DAC, three comparators and a digital controller. The output voltage precision is within 0.6% with universal input voltage.