陳添福

Analog Circuit Design Automation with an Interpretable Knowledge-Guided LLM Planner

資訊科學與工程研究所

類比電路自動化設計、元件尺寸最佳化、gm/ID 設計方法、大型語言模型、可解釋人工智慧、知識萃取、analog design automation、transistor sizing、gm/ID methodology、large language models、explainable artificial intelligence、knowledge extraction

InfoWeaver: An Artifact-Centric Framework for LLM-Based Knowledge Access over Heterogeneous Enterprise Dark Data

網路工程研究所

暗資料、檢索增強生成、大型語言模型、異質文件處理、結構化資料檢索、Dark Data、Retrieval-Augmented Generation、Large Language Models、Heterogeneous Document Processing、Structured Data Retrieval

Closing the Model Improvement Loop: An End-to-End Platform for Task-Specific Small Language Models Adaptation

資訊科學與工程研究所

大型視覺語言模型、端到端模型調適平台、工作流程整合、產物追溯、失敗驅動改進、Large Vision-Language Models、End-to-End Model Adaptation Platform、Workflow Integration、Artifact Traceability、Failure-driven Improvement

Tiling-Aware Latency Simulation for Iteration-Level Scheduling in Large Language Models

資訊科學與工程研究所

大型語言模型、區塊化、設計空間探索、模擬、分析模型、Large Language Models、Analysis Model、Simulation、Design Space Exploration、Tiling

From Representation Learning to Interactive Intelligence: Advanced Time Series Analysis for Industrial Systems

資訊學院博士班・共同指導：彭文志

工業時間序列、表徵學習、大型語言模型、互動式分割、代理人工智慧、不規則數據、Industrial Time Series、Representation Learning、Large Language Models、Interactive Segmentation、Agentic AI、Irregular Data

XMAN: Cross-context Multi-Agent Nexus for Complex Reasoning on Edge

前瞻半導體研究所

地端運算、多模型協作、任務拆解、邊緣計算、小型語言模型、多代理人協作、任務分解、長文件問答、Edge Computing、Multi-Agent Collaboration、Task Decomposition、Small Language Models、Long-Document Question Answering

Code Auto-Completion via Highly Efficient Graph-Based Retrieval Augmented Generation

資訊科學與工程研究所

程式碼檢索、圖結構、增強生成、程式碼自動補全、多語言、靜態分析、MinHash-LSH、code retrieval、graph structure、retrieval-augmented generation、code auto-completion、multi-language、static analysis

Integrating Domain-Specific Task SOPs to Conversational Agents with Ophthalmology Expertise

智能系統研究所

眼科醫療、對話代理、標準作業流程、半自動化生成、資料驗測、Ophthalmology、Conversational Agent、Standard Operating Procedures (SOPs)、Semi-automated conversation generation、Data validation

AutoQA-Gen: Automated QA Dataset Generation and Quality Control from Method Design to System Integration

資訊科學與工程研究所

問答資料集、自動生成、品質控制、知識圖譜、資料增強、代理模型、基準測試、QA dataset、automatic generation、quality control、knowledge graph、data augmentation、proxy model、benchmarking

Pathors: A Path-Based Framework for Building Conversational AI Agents

資訊科學與工程研究所

對話代理、大型語言模型、有限狀態機、中斷處理、對話流程控制、Conversational Agent、Large Language Model、Finite-State Machine、Interrupt Handling、Conversation Flow Control

A Self-Learning Method for Object Detection Based on Mixed Pseudo Labels and Dataset Pruning

資訊學院碩士在職專班・共同指導：郭峻因

深度學習、半監督式學習、混合偽標籤、數據增強、物件偵測、資料剪枝、Deep Learning、Semi-supervised Learning、Pseudo-labeling、Data Augmentation、Object Detection、Data Pruning

Enhancing Table Data Comprehension for Large Language Models: A Preprocessing Framework for Universal Application

資訊科學與工程研究所

大型語言模型、表格數據、數據解析、PDF、Large Language Models、Tabular Data、Data Parsing

Balancing Computation and Data Latency for KV Cache on Edge Devices

資訊科學與工程研究所

大型語言模型、量化、壓縮、Large Language Model (LLM)、KV Cache、Quantization、Compression

Adaptive Deep Hybrid Resolution Network for Precise Time Series Segmentation with Labeling

資訊科學與工程研究所

人工智慧、混合分辨率網路、時間序列分割、時間資料分析、智慧製造、Hybrid Resolution Network、Artificial Intelligence、Time Series Segmentation、Temporal Data Analysis、Industrial Manufacturing

Automated Corpus Inspection and Optimization for Fine-Tuning GPT Models Applied in the RISC-V Domain

網路工程研究所

資料標準化、深度學習微調、RISC-V、大型語言模型、Transformer、Data Standardization、Deep Learning、LLM

Enhancing Large Vision Language Model Understanding through Mixture of Vision Experts

網路工程研究所

多模態大型語言模型、專家混合、視覺指令調校、輕量化微調、視覺編碼器、Multimodal Large Language Model、Mixture-of-Expert、Visual Instruction Tuning、Parameter-Efficient Fine-Tuning、Vision Encoder

Hierarchical RAG for Resolving Global and Local Queries with Local LLMs

智能系統研究所

檢索增強生成(RAG)、全局查詢、本地查詢、大型語言模型 (LLMs)、長上下文模型、層次化資訊整合、微調、合成數據集生成、記憶體優化、成本效益訓練、Retrieval-Augmented Generation (RAG)、Global Query、Local Query、Large Language Models (LLMs)、Long-Context Models、Hierarchical Information Integration、Fine-Tuning、Synthetic Dataset Generation、Cost-Efficient LLM Training

Multi-branch Knowledge Distillation on Tiny Tree-shaped Network for Distributed Co-Inference and Multi-task Learning

資訊科學與工程研究所

微型機器學習、知識蒸餾、行人重辨識、多任務學習、TinyML、Knowledge Distillation、Person-Re-Identification、Multi-task Learning

Complex Embeddings with Relative Position Representations: Rethinking Complex Positional Encoding in Transformer

資訊科學與工程研究所

複數嵌入層、位置編碼、相對位置表示、自注意力機制、Complex embedding、Positional Encoding、Relative Position Representations、Transformer、Self-attention Mechanism

Enhancing Knowledge Distillation in BERT Model: Attention-Guided Distillation and Layer Selection by Attention Diversity

網路工程研究所

神經網路模型壓縮、知識蒸餾、自然語言任務、自注意力模型、model compression、BERT、knowledge distillation、natural language tasks、self-attention

Design Space Exploration for Scalable DNN Accelerators Using a Memory-Centric Analytical Model for HW/SW Co-Design

資通安全碩士學位學程

深度學習、AI加速器、設計空間探索、軟硬體協同設計、硬體模擬、Deep Learning、AI accelerator、Design Space Exploration、HW/SW Co-design、hardware simulator

Efficient Workload Analysis and Traffic Optimization for AI Accelerators by Data Reuse and Traffic Pattern Techniques

網路工程研究所

AI 加速器、分析模型、深度神經網路、工作負載分析、流量優化、AI Accelerator、Analytical Model、Deep Neural Network、Workload Analysis、Traffic Optimization

Efficient Inference of Transformers on Bare-Metal Devices: A Design Flow for Optimizing Models Using RISC-V Vector Intrinsic

資訊科學與工程研究所

裸金屬(bare-metal)、深度學習編譯器、bare-metal、RISC-V、Deep Learning Compiler、Transformer、TVM

Multi-SPCBERT: Time Series Anomaly SPC Patch BERT Serving as an Expert Assistant in Smart Manufacturing

網路工程研究所

統計製程管制、自注意力算法、歷史資料分析、異常偵測、異常時間點定位、SPC control charts、self-attention、history analysis、time series anomaly detection

Efficient Low-bit Quantization for Vision Transformer

資訊科學與工程研究所

視覺型Transformer、模型量化、模型壓縮、知識蒸餾、混合精度模型、Vision Transformer、Quantization、Model Compression、Knowledge Distillation、Mixed-precision Model

Explainable Mask for Root Cause Analysis and Prediction Between FDC / WAT

資訊科學與工程研究所

根因分析、可解釋性神經網路、晶圓預測、高維度資料處理、合成產線資料生成、Root cause analysis、Explainable AI、wafer prediction、high-dimensional data processing、synthesis data generation

Causality Identify via Bayesian Inference and Explainable DNN Prediction

資電亥客與安全碩士學位學程

貝氐網路、因果推論、高維度資料預測、可解釋神經網路、Bayesian networks、Causal inference、High-dimensional data、Explainable Artificial Intelligence

A Multi-Head Ensemble Multi-Task Learning Framework for RUL Prediction and Health Status Assessment

資訊科學與工程研究所

剩餘使用壽命預測、健康狀況評估、多任務學習、知識蒸餾、集成、RUL prediction、Health status assessment、Multi-task learning、Knowledge distillation、Ensemble

An Incremental Optimization Method for High Dimensional Multi-Objective Problems with Prior Knowledge

資訊科學與工程研究所

參數最佳化、黑盒函數最佳化、高維度最佳化問題、多目標最佳化、先備知識、black-box optimization、high-dimensional black-box optimization、multi-objective optimization、optimization with prior knowledge

Divide-and-Conquer Online Knowledge Distillation and Quantization-Aware Training by Dynamic Blending

資訊科學與工程研究所

知識蒸餾、量化訓練、Knowledge Distillation、Quantization-Aware Training

Enhancing Semi-Supervised Learners via Similarity-Based Self-Supervised Models

資訊科學與工程研究所

半監督式學習、自監督式學習、表征學習、知識蒸餾、異常數據偵測、Semi-Supervised Learning、Self-Supervised Learning、Representation Learning、Knowledge Distillation、Out-of-Distribution Detection

Adaptive Similarity-Aware Hyperparameter Tuners for Classification Tasks

資訊科學與工程研究所

超參數搜尋、遷移學習、分類問題、資料集相似度、演化計算、Hyperparameter Tuning、Transfer Learning、Classification、Dataset Similarity、Evolutionary Algorithm

Low-bit Backpropagation Algorithm Design for Learning on Edge Devices

資訊科學與工程研究所

神經網路優化、量化、低精度訓練、量化訓練、終端學習、邊端學習、Fully quantized training、Low-bit training、On-device learning、Neural networks、Quantization

Learning Deep Representation and Conditional Probability Distribution for Outlier Exposure Open Set Recognition

資訊科學與工程研究所

開放式識別、異常檢測、條件機率、Conditional Probability Distribution、Open set recognition、Outlier exposure

Optimizing Neural Network Compilation via Adaptive Workflow with Shared History Data

資訊科學與工程研究所

強化學習、神經網路、編譯器、Reinforcement Learning、Neural Networks、Compilers

Hybrid Bayesian Optimization with Early Termination Strategies for Auto Tuning

電機資訊國際學程

人工智能、超參數調整、Artificial Intelligence、Hyperparameter Tuning、Algorithm Scheduling、Early Termination、Edge devices

Quantization and Pruning with Differentiable Masks

資訊科學與工程研究所

剪枝、quantization

Data-Free Target-Aware Model Compression System for Efficient Parallel Inference

資訊科學與工程研究所

免數據集、模型壓縮、平行推論、data-free、model compression、parallel inference

Target and Dataset Oriented Neural Network Specialization

資訊科學與工程研究所

深度學習、類神經網路、落地問題、基於目標平台、基於目標資料集、類神經網路搜尋、遷移訓練、Deep Learning、Neural Network、Landing Problem、Target Device Aware、Target Dataset Aware、Neural Architecture Search、Transfer Learning

Adaptive Tuner for Task-aware Hyperparameter Tuning

資訊科學與工程研究所

超參數調整、深度學習、強化學習、Hyperparameter Tuning、Deep Learning、Reinforcement Learning

Target-aware Neural Architecture Search Coupled with Channel Pruning

資訊科學與工程研究所

類神經網路搜索、通道剪枝、Neural Architecture Search、Channel pruning、NAS

Compile Deep Neural Networks to NVDLA Using Glow

資訊科學與工程研究所

深度學習、加速器、編譯器、Deep learning、accelerator、compiler

Fine-Grained Neural Network Profiler for Target-aware Optimization Flow Improvement

資訊科學與工程研究所

分析器、深度學習、動態二維插點、deep learning、profiler、dynamic binary instrumentation

Efficient 4-bit & 8-bit Quantization for CNNs without Retraining

資訊科學與工程研究所

四位元、卷積神經網路、量化、4-bit、Quantization、CNNs

Separate Quantization with Dual Precisions for Convolutional Neural Networks

資訊科學與工程研究所

卷積神經網絡、量化、降低精度、convolutional neural networks、quantization、reduced precision

Target-aware Optimization Mechanism for Convolutional Neural Networks

資訊科學與工程研究所

深度學習、卷積神經網路、模型壓縮、超參數優化、Deep Learning、Convolutional Neural Networks、Model Compression、Hyper-parameter Tuning

A Sparse Optimization Design for Pointwise Convolutions on the CNN Accelerator

資訊科學與工程研究所

物聯網、機器學習、加速器、IoT、Machine Learning、Accelerator

Sensor Abnormal Detection and Recovery in IoT Sensing System Based on Machine Learning

資訊科學與工程研究所

感知器、機器學習、錯誤偵測、錯誤修復、物聯網、sensor、IoT、machine learning、fault recovery、fault detection

Memory Architectures for Embedded Binarized Neural Networks on IoT Devices in Low Voltage Environment

資訊科學與工程研究所

低電壓、物聯網、嵌入式系統、人工神經網路、預存取、暫存器緩存、Low voltage、Internet of Things、IoT、Embedded system、Prefetch、Register buffer

Write-Hotness Aware Cache-Set Remapping for Improving Lifetime of Non-volatile Caches

資訊科學與工程研究所

非揮發性快取、耗損平衡、non-volatile cache、wear leveling

Optimizing Caffe Training Framework for Gateway Systems

資訊科學與工程研究所

閘道器系統、深度學習、傳導卷積神經網路、效能最佳化、Gateway systems、Caffe、CNN

Reducing Timing Discrepancy for Energy-Efficient On-Chip Memory Architectures with Low Voltage Processors

資訊科學與工程研究所

低電壓處理器、錯誤容忍快取記憶體、降低時序差異、Low-voltage processor、Fault-tolerant cache、Timing discrepancy reducing

An Energy-Efficient and Performance-Optimized Memory Architecture in Multicore Systems

資訊科學與工程研究所

多核心、探聽過濾器、硬體預取、服務品質、排程器、記憶體控制器、multicore、snoop filter、hardware prefetching、QoS、scheduler、memory controller

System Architecture Optimizations for Energy-Efficiency with Emerging Memory Systems

資訊科學與工程研究所

新興記憶體、系統架構優化、能源效率、emerging memory、system architecture、energy efficiency

A Feature-Rich and Energy-Efficient Regular Expression Matching Accelerator via Non-Volatile Memory Architecture

資訊科學與工程研究所

正規表示式比對、優先權編碼器、末級快取記憶體、深度封包檢測、非揮發性靜態隨機存取存儲、Regular expression matching、Priority encoder、Last level cache、Deep packet inspection、Non-volatile SRAM

Garbage Collection and Memory Management Optimization in R Language for IoT Gateway System

資訊科學與工程研究所

物聯網、物聯網閘道系統、R語言、記憶體回收、平行化、記憶體管理、IoT、IoT Gateway System、R language、Garbage collection、Parallel、Memory management

Mezzo-Grained DRAM Cache with Main Memory Speculative Access Command

電機資訊國際學程

主記憶體、隨機存取存儲器、DRAM Cache、Dual Granularity、Speculative Access

Power Evaluation Platform for Low Voltage IoT Architecture and Its Usage in Energy-Efficient Design

資訊科學與工程研究所

物聯網、功率評估平台、IoT、Power evaluation platform、Energy efficient

Scalable Fault Tolerant IoT Coordinated Gateway System Based on Distributed Service

資訊科學與工程研究所

物聯網、物聯網閘道器、容錯能力、分散式服務、IoT(Internet of Things)、IoT gateway、Fault Tolerance、Distributed Service、Docker

TCAM-based Low Power Packet Classification Module and Wildcard Pattern Matching Module in DPI Systems

資訊科學與工程研究所

封包分類、字串比對、packet classification、Content Inspection

Configurable Profiling Tool and Evaluation Platform for Low-End Embedded Cores

資訊科學與工程研究所

異質多核心、穿戴式裝置、模擬平台、可調式、Heterogeneous、Wearable device、Simulation platform、Configurable

A Full-system Heterogeneous Multicore Simulation Platform and Accelerating Architecture-Level Simulation by Sampling Simulation

資訊科學與工程研究所

系統架構模擬平台、取樣模擬加速、異質多核心、Dynamic adjustment and partial sampling simulation、Multithreaded/Multicore simulation、GPGPU simulation、HSA simulation

Leak Stopper: An Actively Revitalized Snoop Filter Architecture by Effective Generation Filtering

資訊科學與工程研究所

快取儲存、記憶體架構、過濾器、計算機架構、Cache storage、memory architecture、filters、computer architecture

A Full System Simulation Framework for HSA and CTA-based Prefetch Mechanism

資訊科學與工程研究所

異質多核心、模擬、協作緒陣列、資料預取、Simulation、hetergeneous、HSA、prefetch

Improving DRAM Cache Hit Rate and Performance via Adaptive Granularity Block Size Management

資訊科學與工程研究所

快取記憶體、粒度、自適、命中率、DRAM Cache、Granularity、Adaptive、Hit Rate

A Study of System Behavior Impacts on Memory Hierarchy Exploration Based on Full-System Simulation

資訊科學與工程研究所

系統、模擬、記憶體階層、效能分析、full-system、simulation、memory hierarchy、profiling

Word-line Alliable Cache for Instantaneous Performance Boost on Asymmetric Core Systems

資訊科學與工程研究所

低電壓快取記憶體設計、非對稱式處理器架構、靜態隨機存取記憶體、low voltage cache design、asymmetric core systems、SRAM

Vertical profiling monitor and fine-grained resource provision for virtualized server management

資訊科學與工程研究所

虛擬化伺服器、垂直分析、精細資源分配、virtualized server、Vertical profiling、fine-grained resource provision

A Self-Adaptive Mapping Scheme to Assist Page Allocation for DRAM Energy Efficiency

資訊科學與工程研究所

記憶體、位址分配、低功耗、DRAM、address allocation、low power

MMU Cache System and Thread Block Scheduling Enhancement for Virtual Memory Support on GPGPU

資訊科學與工程研究所

圖形處理單元、異質運算架構、記憶體管理單元、虛擬記憶體位置轉換、GPU、HSA、MMU、Virtual address translation

Sampling Simulation through Program Characteristic Analysis for Modern Multicore Architectures

資訊科學與工程研究所・共同指導：徐慰中

模擬、取樣、圖形處理器、程式特性、simulation、sampling、GPU、program characteristic

A Fine-Grained Line-Mergeable Cache to Reduce Timing Discrepancy with Low Voltage Processors

資訊科學與工程研究所

低電壓快取記憶體、列合併快取記憶體、降低時序差異、Low voltage cache、Line-mergeable cache、Reduce timing discrepancy

Multi-granularity Vertical Profiling Tool with Backtrace for Android Systems

資訊學院資訊學程

安卓、多重細度、回溯、垂直分析、虛擬機器、嵌入式系統、Android、Multi-granularity、Vertical Profiling、Backtrace、Embedded System

Elastic Cache: Leverage Latency to Minimize Energy on Low-Voltage Fault-Tolerant Cache

資訊科學與工程研究所

低電壓快取記憶體、容錯、製程變異、能源效率、可靠度、延遲察覺、Low Voltage Cache、Fault-Tolerant、Process variation、Energy efficiency、Reliability、Latency aware

Lifetime-aware LRU Promotion Policy for Last-level Cache

資訊科學與工程研究所

末級快取、提升機制、生命週期、上層快取資訊、輾轉現象、多核心、Last-level cache、Promotion policy、Lifetime-aware、high-level cache hint、thrashing、Multicore

Adaptive Granularity and Coordinated Management for Timely Prefetching in CMPs

資訊科學與工程研究所

預取、節流、快取、多核心、prefetch、throttle、cache、CMP、thread correlation

DAPs : Dynamic Adjustment and Partial Sampling for Multi-threaded / Multi-core Simulation

資訊科學與工程研究所

模擬器、取樣、多執行緒、Simulation、Sampling、Multi-threaded

Adaptive Prefetching Techniques and Latency Overlapping Scheduling for 3D Wide I/O Memory

資訊科學與工程研究所

寬頻記憶體、調適提前讀取技術、重疊延遲排程、Wide I/O Memory、Adaptive Prefetching Techniques、Latency Overlapping Scheduling

Embedded Debugging Architecture of Bug Detection and Reproduction for SoC Post-Silicon Validation

資訊科學與工程研究所

後矽驗證、錯誤偵測、錯誤重現、除錯架構、SoC Validation、Error Detection、Error Reproduction、Debugging Architecture

Leveraging Adaptive Set-Configurable Architecture to Reduce Conflict Misses of PCM Last-Level Cache in CMP

資訊科學與工程研究所

快取架構、單晶片多處理器、衝突性失誤、末級快取、變相化記憶體、cache architecture、chip multiprocessor、conflict miss、last-level cache、phase change memory

Low Overhead Context-sensitive Vertical Profiling for Virtual Machine Based Mobile Embedded Devices

資訊學院資訊學程

垂直性能分析、Dalvik虛擬機器、Java、Android、記憶體通道、Dalvik插入測試模組、Vertical Profiling、Dalvik Virtual Machine、Direct Memory Tunnel、Dalvik Instrumentation Module