https://scholar.dgist.ac.kr/handle/20.500.11750/12956 2026-08-17T04:07:32Z https://scholar.dgist.ac.kr/handle/20.500.11750/60479 Title: A Hybrid Digital-Analog Compute-in-Memory Using Content-Addressable Memory With Flexible Multi-Bit Slicing Author(s): Jung, Sangwoo; Lee, Hojin; Park, Jiyong; Lee, Yejin; Park, Dahoon; Shin, Hyunseob; Yoon, Jong-Hyeok; Kung, Jaeha Abstract: Compute-in-memory (CIM) reduces data movement and enhances compute parallelism, making it suitable for AI applications. However, analog CIMs, yet energy-efficient, are vulnerable to PVT variations, while digital CIMs offer robustness but limited efficiency due to their bit-wise computation overhead. To address these challenges, we propose a hybrid CIM architecture that integrates content-addressable memory (CAM) and cluster-based CIM, named CAM-CIM, fabricated in 65nm CMOS technology. The proposed CAM-CIM flexibly slices multi-bit weights, assigning MSBs to CAM and LSBs to CIM, enabling dynamic accuracy-efficiency trade-offs across various bit precisions. A two-stage 8:3 compressor-based adder tree improves CAM efficiency and a reference voltage search algorithm ensures accurate CIM computation with low-bit ADCs. Our CAM-CIM supports 1-8b inputs/weights with reconfigurable compute modes, leveraging ternary-CAM based selective columns and cluster-wise CIM processing to produce multiple trade-off points even in the same bit precision. A prototype chip with a RISC-V controller and custom instructions is demonstrated that shows energy efficiencies of 32.4TOPS/W (8b/8b) and 76.0-354.9TOPS/W (4b/4b) with 0.66% accuracy loss, on average, across a wide range of DNN benchmarks including CNNs and vision transformers on CIFAR and ImageNet datasets. 2026-04-30T15:00:00Z https://scholar.dgist.ac.kr/handle/20.500.11750/60122 Title: CAM-CIM: A Hybrid Compute-in-Memory Using Content-Addressable Memory with Subword Split Mapping for Reduced ADC Resolution Author(s): Jung, Sangwoo; Lee, Hojin; Lee, Yejin; Park, Jiyong; Park, Dahoon; Shin, Hyunseob; Yoon, Jong-Hyeok; Kung, Jaeha Abstract: Recently, compute-in-memory (CIM) has become a promising architecture for data-intensive applications such as deep learning. However, analog or digital CIM (ACIM or DCIM) faces some design challenges. ACIMs inherently have non-idealities, which lead to significant accuracy degradation. In addition, a substantial amount of power is consumed by analog-to-digital converters (ADC). On the other hand, DCIMs show an exponential increase in power consumption and computing cycles as the operand bit-width increases, particularly due to an accumulation stage. In this paper, to overcome these challenges, we propose a hybrid DCIM-ACIM architecture that consists of a content addressable memory (CAM) as DCIM and a cluster-based multi-cycle ACIM, called CAM-CIM. As a weight mapping strategy, we present a subword split mapping that assigns some MSBs to DCIM for improved accuracy and the remaining LSBs to ACIM for reduced ADC resolution. The accuracy of using the proposed CAM-CIM array is evaluated on various deep learning benchmarks from CNNs to Swin-Tiny. A 65nm CAM-CIM macro with either 3-bit or 4-bit ADCs shows 10.3x and 5.4x improvement in energy efficiency, on average, compared to CAM- and CIM-only architectures, respectively. Compared to recent CIM architectures, CAM-CIM demonstrates 1.4x higher energy efficiency. 2025-08-07T15:00:00Z https://scholar.dgist.ac.kr/handle/20.500.11750/59973 Title: High-Spatiotemporal-Resolution Transparent Thermoelectric Temperature Sensor Arrays Reveal Temperature-Dependent Windows for Reversible Photothermal Neuromodulation Author(s): Lee, Junhee; Yoon, Dongjo; Lee, Jungha; Kim, Duhee; Kim, Eunui; Yoon, Jong-Hyeok; Kwon, Hyuk-Jun; Chung, Seungjun; Nam, Yoonkey; Kang, Hongki Abstract: Photothermal neural stimulation enables optical excitation or inhibition of neural activity depending on the dynamics of localized temperature changes, offering high spatial resolution without genetic modification. However, quantitative analysis of these temperature dynamics remains limited due to the lack of suitable direct sensing technologies, posing a challenge to the safe and controlled application of photothermal neural stimulation techniques. This challenge is addressed by developing transparent thermoelectric temperature sensor arrays with high spatiotemporal resolution, integrated with electrical and optical recording capabilities. These microscale sensors stably and accurately capture rapid temperature increases and decreases, and thermal equilibrium induced by thermo-plasmonic effects at the neural interface, regardless of the environment. The multifunctional platform allows simultaneous electrical and optical monitoring of neural responses during the photothermal stimulation, enabling detailed analysis of the correlation between localized temperature changes and neural activities. a reversible neural inhibition window (1.4-4.5 degrees C) and thresholds for irreversible damage (>6.1 degrees C) are identifyed. Using high temporal-resolution sensing, localized thermo-plasmonic temperature dynamics over tens of milliseconds, and associated neural signal suppression and reactivation are captured. This approach provides unprecedented insight into the interplay between photothermal effects and neural activity, establishing a foundation for precise, temperature-guided neuromodulation therapies and advanced neural circuit research. 2026-01-31T15:00:00Z https://scholar.dgist.ac.kr/handle/20.500.11750/59929 Title: Translational reprogramming of dentate gyrus peptidergic circuitry gates antidepressant efficacy Author(s): Oh, Seo-Jin; Jang, Jin-Hyeok; Roussarie, Jean-Pierre; Jang, Kyung-un; Jeong, Min-Seok; Jo, Yeon Suk; Shin, Chang Hun; Choi, Hongsoo; Lee, Kwang; Yoon, Jong-Hyeok; Oh, Yong-Seok Abstract: Selective serotonin reuptake inhibitors (SSRIs) exhibit delayed therapeutic effects despite rapid serotonin elevation, suggesting their dependence on slow neuroplastic adaptations. Here, we demonstrate that antidepressant actions require cell type-specific translational regulation of the peptidergic signaling in the dentate gyrus (DG). Chronic, but not acute, treatment with an SSRI fluoxetine (FLX) selectively enhances translational activity in hilar mossy cells (MCs), with no detectable changes in neighboring granule cells (GCs). Combining Translating Ribosome Affinity Purification (TRAP) with RNA sequencing revealed distinct baseline translatomes between these two glutamatergic neurons and identified FLX-induced remodeling of peptidergic pathways in the DG. Crucially, we discovered MC-specific enrichment of the neuropeptide PACAP, which undergoes translation-dependent upregulation by chronic FLX treatment. This PACAP induction mediates neuroadaptive plasticity in PAC1 receptor-expressing GCs and drives behavioral responses prominently in female mice during prolonged FLX administration. Our findings establish cell type-specific translational reprogramming as a novel mechanistic framework for antidepressant action. 2026-01-31T15:00:00Z