Zhaori Bi (毕朝日)

Numeric optimization methods are widely utilized to tackle complex analog circuit sizing problems, where the challenges include expensive simulations, non-linearity, and high parameter dimensionality. However, the diverse characteristics exhibited by different circuits result in varied optimization landscapes, making it difficult to identify a single algorithm that consistently outperforms others across all problems. This paper proposes MARIO, a multi-algorithm interworking optimization framework that achieves optimization superadditivity based on a pool of member algorithms and a powerful algorithm-interworking protocol. A computing resource reallocation method is proposed, which employs multi-task Gaussian process regression and portfolio optimization techniques, leading to flexible and prudent online adaption of member algorithms. Furthermore, a cooperative local refinement feature is introduced, allowing algorithms to share evaluation data and leverage each other’s strengths for local search. Experimental results on both synthetic benchmarks and real-world analog circuits demonstrate that MARIO surpasses existing algorithm-combining tools and standard optimization methods.

Automated analog circuit design migration significantly alleviates the burden on designers in circuit sizing under various operating conditions. Conventional methods model the migration problem as black-box optimization, requiring excessive iterations of costly simulations to converge. Reinforcement learning exhibits significant promise in transfer learning, as it enables the generation of circuits that fulfill specifications efficiently. This paper proposes EVDMARL, a novel value decomposition-based multi-agent reinforcement learning framework that models complex analog circuits and eliminates the need for manually defined specifications of sub-circuits for new operating conditions. Furthermore, domain randomization is incorporated to efficiently generate circuits that meet unforeseen scenarios with minimal simulations. Experimental results on real-world analog circuits demonstrate the effectiveness of the proposed framework for complex analog circuit design migration.

A constrained Voronoi tree-based domain decomposition method for high-dimensional Bayesian optimization is proposed to solve large scale analog circuit synthesis problems, which can be formulated as high-dimensional heterogeneous black-box optimization. The hierarchical Voronoi tree progressively breaks down the design space into partitions with implicit performance boundaries such that promising regions are efficiently explored. Combined with a local Bayesian optimization strategy within each partition, the method effectively balances exploration and exploitation in high-dimensional spaces. Experimental results on both synthetic benchmarks and real-world analog circuits demonstrate that the proposed method significantly outperforms state-of-the-art methods in terms of optimization quality and convergence speed.

High-dimensional analog circuit synthesis is a challenging simulation-expensive optimization problem. This paper proposes BBGP-sDFO, a novel batch Bayesian and Gaussian process enhanced subspace derivative free optimization method to solve high-dimensional and simulation-expensive analog circuit optimization problems. The method includes a batch Bayesian querying strategy for exploring the global design space and a Gaussian process enhanced subspace derivative free optimization method for exploiting promising regions in effective low-dimensional subspace. By combining global exploration with local exploitation in reduced subspaces, BBGP-sDFO efficiently handles the curse of dimensionality while maintaining high optimization quality. Experimental results on both synthetic benchmarks and real-world analog circuits demonstrate that BBGP-sDFO significantly outperforms state-of-the-art methods in terms of optimization efficiency and solution quality for high-dimensional analog circuit synthesis.

Knee osteoarthritis (OA) is a chronic disease that considerably reduces patients’ quality of life. Preventive therapies require early detection and lifetime monitoring of OA progression. In the clinical environment, the severity of OA is classified by the Kellgren and Lawrence (KL) grading system, ranging from KL-0 to KL-4. Recently, deep learning methods were applied to OA severity assessment to improve accuracy and efficiency. However, this task is still challenging due to the ambiguity between adjacent grades, especially in early-stage OA. Targeting the uncertainty in the OA dataset, we propose a novel learning scheme that dynamically separates the data into two sets according to their reliability. Besides, we design a hybrid loss function to help CNN learn from the two sets accordingly. With the proposed approach, we emphasize the typical samples and control the impacts of low confident cases. Our method achieves a mean accuracy of 70.13% on the five-class OA assessment task, which outperforms all other state-of-art methods. Moreover, we design an experiment to validate large-scale automatic data refining during training. The dataset used in this paper was obtained from the Osteoarthritis Initiative.

Hearing loss is one of the most common modifiable factors associated with cognitive and functional decline in geriatric populations. An accurate, easy-to-apply, and inexpensive hearing screening method is needed to detect hearing loss in community-dwelling elderly people. In total, 1793 participants aged 60 years and above were recruited to undertake a standard audiometric air conduction pure tone test at 4 frequencies. The decision tree analysis identified 2 kHz as the most important frequency for moderate-to-profound hearing loss screening in elderly people. The tree model suggested a simple two-step screening procedure in which a 42 dB HL tone at 2 kHz is presented first, followed by a 47 dB HL tone at 0.5 kHz, depending on the individual’s response to the first tone. This approach achieved an accuracy of 91.20%, a sensitivity of 95.35% and a specificity of 86.85% in the training dataset. The two-tone screening procedure provides an effective and practical method for identifying moderate-to-profound hearing loss in community-dwelling older adults.