簡仁宗

Information-Theoretic Distillation and Compression via Flow Alignment

電機工程學系

大型語言模型壓縮、知識蒸餾、前饋網路參數共享、自適應提前退出、Large language model compression、Knowledge distillation、Feed-forward parameter sharing、Adaptive early exit

Inference-Time Derisking in Masked Diffusion

資訊安全研究所

擴散語言模型、安全生成、推論時引導、Masked Diffusion Language Model、Safety Generation、Inference-Time Guidance

Temporal-Coupled Flow Matching for Speech Continuation

電信工程研究所・共同指導：王忠炫

語音延續、時間耦合流匹配、語音語言模型、端點軌跡、有限差分正則化、參數高效率調適、Speech Continuation、temporal-coupled flow matching、spoken language model、endpoint trajectory、finite-difference regularization、parameterefficient adaptation

Physics-Inspired Landscape Learning and Flow Matching

電信工程研究所・共同指導：王忠炫

結構式藥物設計、流匹配、殘差控制、軌跡層級控制、蛋白質-配體 生成、物理啟發式導引、Structure-Based Drug Design、Flow Matching、Residual Control、Trajectory-Level Control、Protein–Ligand Generation、Physics-Inspired Guidance

Variational Optimal Transport for CTC Alignment

電信工程研究所

連線式時序分類對齊、最佳傳輸、變分推論、音框--詞元指派、後驗尖峰化、自動語音辨識、CTC alignment、Optimal Transport、variational inference、frame-token assignment、posterior peakiness、automatic speech recognition

Bidirectional ASR-TTS Flow Matching

電機工程學系

自動語音辨識、語音合成、統一機率模型、雙向流匹配、Speech recognition、Speech synthesis、Unified probabilistic modeling、Bidirectional flow matching

State-Space Flow Matching for Shortcut Generation

電機工程學系

擴散模型、流匹配、文字到影像生成、少步生成、知識蒸餾、狀態空間模型、硬體友善推論、Diffusion Models、Flow Matching、Text-to-Image Generation、Few-Step Generation、Knowledge Distillation、State-Space Model、Hardware-Friendly Inference

Diffusion Learning via Strategic Denoising and Remasking

電機工程學系

離散擴散模型、遮罩式擴散語言模型、重新遮罩、去噪、策略最佳化、軌跡展開學習、Discrete diffusion model、Masked diffusion language model、Remasking、Denoising、Policy optimization、Rollout-based learning

Learning Towards Unmasking and Remasking by Actor-Critic Diffusion

電機工程學系

離散擴散模型、遮罩式擴散語言模型、重新遮罩、推理修正、策略最佳化、軌 跡展開學習、Discrete diffusion model、Masked diffusion language model、Remasking、Reasoning repair、Self-correction、Policy optimization、Mathematical reasoning

Convolutional State-Space Propagation for Speech Recognition

電信工程研究所

自動語音辨識、狀態空間模型、卷積、大型語言模型、automatic speech recognition、state space model、convolution、large language model

Adversarial Contrastive Learning for Efficient RAG

電機工程學系

大型語言模型、檢索增強生成、提示壓縮、對比學習、對抗訓練、Large language model、retrieval-augmented generation、prompt compression、Contrastive Learning、Adversarial Training

Learnable Contrastive Decoding in State-Space LLMs

資訊學院碩士在職專班

大型語言模型、狀態機、幻覺、對比式解碼、模型編輯、早期退出、Large language model、state space machine、hallucination、contrastive decoding、model editing、early exit

Editing Mixture of Experts for Speech Synthesis

電信工程研究所・共同指導：王忠炫

多語言語音合成、專家混合架構、知識編輯、推理效率、Multilingual Text-to-Speech、Mixture of Experts、Knowledge Editing、Inference Efficiency

Learning Corrective Memory at Test Time

電信工程研究所・共同指導：王忠炫

序列建模、狀態空間模型、記憶更新、測試時訓練、語意校正、sequence modeling、state space models、memory update、test-time training、semantic correction

Temporal Alignment in Attentive State-Space Models for Speech Recognition

人工智慧技術與應用碩士學位學程

自動語音識別、連結時序分類、狀態空間模型、自回歸生成、單調對齊搜索、自注意力機制、automatic speech recognition、end-to-end speech recognition、connectionist temporal classification、state spaces model、self-attention、autoregressive generation、monotonic alignment search

Trilevel Supervised, Unsupervised, and Distilled Learning for Speech Recognition

電機工程學系

自動語音辨識、三層最佳化、監督式學習、非監督式學習、知識蒸餾、automatic speech recognition、trilevel optimization、supervised learning、unsupervised learning、knowledge distillation

Constraint-Based Reward Tuning (CBRT): Escaping Sub-Optimal Regions in Reinforcement Learning via Landscape-Guided Reward Adjustment

人工智慧技術與應用碩士學位學程

強化學習、獎勵調整、對偶方法、自適應約束、次優陷阱、Reinforcement Learning、Reward Adjustment、Primal-Dual Methods、Adaptive Constraints、Sub-Optimal Traps

Log-Normal State-Space Model

電機工程學系

線性注意力、狀態空間模型、對數常態分佈、長距離依賴、序列建模、Linear attention、State-space model、Log-normal distribution、Long-range dependency、Sequence modeling

Optimal Controlled Diffusion for Human Motion Transfer

電機工程學系

電腦視覺、擴散模型、最優控制、生成式視覺模型、影片生成、人體動作轉移、computer vision、diffusion models、optimal control、generative video、human motion transfer

Recognition of Mandarin Tones Using Brain Dynamics for Brain-Computer Interfaces

電機工程學系・共同指導：李鎮宜

腦電波、華語聲調、特徵擷取、華語語音資料庫、有理數化之不對 稱特徵提取方法（RASM）、EEG、Mandarin Tone、Feature extraction、Mandarin Speech Database、rational asymmetric feature extraction method (RASM)

Reinforced Retrieval to Avoid Text Hallucination

電機工程學系

大型語言模型、幻覺、檢索增強生成、強化學習、強化檢索、文本生成、Large Language Models、Hallucination、Retrieval-Augmented Generation、Reinforcement Learning、Reinforced Retrieval、Text Generation

Contrastive Disentanglement Learning for Empathetic Dialogue Generation

電機工程學系

自然語言生成、解纏表示學習、對比學習、提示學習、大型語言模型、Natural Language Generation、Disentangled Representation Learning、Contrastive Learning、Prompt Learning、Large Language Model

Learning Diffusion Transitions for Continuous-Discrete Denoising Process

人工智慧技術與應用碩士學位學程

自然語言生成、擴散模型、掩碼語言模型、簡單優先、高斯分布、狄拉克分布、對比損失、Natural Language Generation、Diffusion Models、Mask Language Models、Easy-First Generation、Gaussian Distribution、Dirac Distribution、Contrastive Loss

Semantic Diffusion Model with Attention Disentanglement for Text-to-Image Generation

電機工程學系

電腦視覺、文本生成圖像模型、圖像生成、擴散模型、多模態學習、解糾纏、互訊息、computer vision、text-to-image model、image generation、diffusion model、multimodal learning、disentanglement、mutual information

Probabilistic Language-Aware Speech Recognition

電信工程研究所・共同指導：王忠炫

自動語音識別、語言切換、參數效率、機率方法、automatic speech recognition、code-switching、parameter-efficient、probabilistic approach

CAPR: Confidence-Aware Prompt Refinement for DeHallucination

資通安全碩士學位學程

信心度引出、信心度校準、大型語言模型、強化學習、提示優化、Confidence elicitation、Confidence calibration、Large Language Model、Reinforcement learning、Prompt refinementiv

Diffusion Model for Non-Autoregressive Speech Recognition

電機工程學系

自動語音辨識、連結時序分類、擴散模型、非自迴歸生成、動態規劃、快速取樣、automatic speech recognition、connectionist temporal classification、diffusion model、non-autoregressive generation、dynamic programming、fast sampling

Knowledge-infused Topic Model for Empathetic Dialogue Response

電機工程學系

共情對話回復、主題模型、圖神經網路、預訓練模型、empathetic dialogue response、topic model、graph neural network、pre-trained model

Variational Dehallucination via Retrieval-Augmented Prompt Learning

電信工程研究所

大型語言模型、幻覺、變分推斷、檢索增強生成、提示學習、互消息、large language model、hallucination、variational inference、retrieval-augmented generation、prompt learning、mutual information

Learning Contrastive Emotional Nuances in Speech Synthesis

電機資訊國際碩士學位學程

情感文本語音合成、對比學習、解纏、Emotional text-to-speech、Contrastive learning、Disentanglement

Reinforcement Learning and Prompting Towards Reliable Dialogue System

電機資訊國際博士學位學程

強化學習、提示策略、可靠的對話系統、語言理解、語言生成、Reinforcement Learning、Prompting Strategy、Reliable Dialogue System、language Understanding、Language Generation

Modality Translation Learning for Joint Speech-Text Model

電信工程研究所・共同指導：王忠炫

自動語音辨識、聯合語音文字模型、跨模態特徵表示、自監督式學習、automatic speech recognition、joint speech-text model、cross-modality representation、self-supervised learning

Attention-Guided Adaptation for Code-Switching Speech Recognition

電機資訊國際碩士學位學程

注意力引導、語碼轉換、參數高效、雙語語音識別、Attention guidance、code-switching、parameter efficiency、bilingual speech recognition

Continual Gated Adapter for Bilingual Neural Codec Text-to-Speech

電機工程學系

神經編解碼語言建模、調適器調整、雙語語音合成、持續學習、neural codec language modeling、adapter tuning、bilingual text-to-speech、continual learning

Collaborative Contrastive Learning for Hypothesis Domain Adaptation

資通安全碩士學位學程

語者驗證、領域適應、協同式學習、對比式學習、speaker verification、domain adaptation、collaborative learning、contrastive learning

Dynamic Mixup Augmentation for Meta Soft Prompt Learning

電機工程學系

少樣本學習、元學習、軟提示、混合式擴增、對比學習、自然語言理解、few-shot learning、meta-learning、soft prompt、mixup、contrastive learning、natural language understanding

Cross-Modality Feature Decorrelation in Discrete Diffusion Model

電信工程研究所・共同指導：王忠炫

自動語音識別、大規模預訓練聲學模型、離散擴散模型、特徵解相關化學習、automatic speech recognition、large-scale pretrained acoustic model、discrete diffusion model、feature decorrelation

Emotional Dialogue Act Selection for Empathetic Generation

電機工程學系

深度學習、序列對序列學習、自監督學習、注意力模型、轉換器、語言模型、自然語言處理、Deep learning、sequence-to-sequence learning、self-supervised learning、attention network、transformer、language model、natural language processing

Empathetic Response Generation via Regularized Q-Learning

電機工程學系

離線強化學習、軟Q學習、共情對話、自然語言生成、offline reinforcement learning、soft Q-learning、empathetic dialogue、natural language generation

Learning Adapters for Code-Switching Speech Recognition

人工智慧技術與應用碩士學位學程

語碼轉換方案、多語言語音辨識、適配器、語言辨識、預訓練模、wav2vec2.0ii、Code-switching scheme、multilingual speech recognitio、adapter、language identification、pre-trained model

Variational Contrastive Learning for Empathetic Dialogue Generation

電信工程研究所

共情對話、變異推論法、對比式學習、預訓練模型、生成模型、empathetic dialogue、variational inference、contrastive learning、pre-trained model、generative model

Learning to Enhance the Text Position Embedding and Coherence

電機工程學系

轉換器、序列式學習、位置編碼、機器翻譯、自然語言處理、Transformer、sequential learning、positional embedding、machine translation、natural language processing

Contrastive Goal Inference for Dialogue Systems

電機工程學系

離線強化學習、對比學習、目標導向強化學習、對話策略、任務導向對話、基於知識之對話系統、offline reinforcement learning、contrastive learning、goal-conditioned reinforcement learning、dialogue policy、task-oriented dialogue、knowledge-based dialogue

Enhancing the Empathetic Dialogues through Emotional Attention Network

電機工程學系

對話系統、共情聊天機器人、共同注意力網絡、語言適配、自然語言生成、dialogue system、empathetic chatbot、co-attention network、language adapter、natural language generation

Learning Outlier-Aware Representation with Synthetic Boundary Samples

電機工程學系

分佈外檢測、對比式學習、表徵學習、電腦視覺、自然語言處理、out-of-distribution detection、contrastive learning、representation learning、computer vision、natural language processing

Domain-Agnostic Soft Prompt via Meta Learning

人工智慧技術與應用碩士學位學程

無監督領域自適應、提示學習、元學習、自然語言處理、自然語言理解、Unsupervised domain adaptation、Prompt-based learning、Meta-learning、Natural language processing、Natural language understanding

Collaborative Pseudo Labeling for Contrastive Language Models

電機工程學系

提示字學習、對比式學習、自監督學習、語言模型、自然語言處理、注意力模型、轉換器、偽標籤、半監督學習、小資料學習、prompt-based learning、contrastive learning、language model、few-shot learning、pseudo-labeling、semi-supervised learning

Adversarial Data Augmentation in Adapter for Language Understanding

電機工程學系

預訓練模型、微調、自然語言處理、參數效率、對抗性訓練、適配器微調、Pre-trained model、fine-tuning、natural language processing、parameter efficiency、adversarial training、adapter

Towards a Unified View of Adversarial Training: A Contrastive Perspective

人工智慧技術與應用碩士學位學程・共同指導：歐陽盟

對抗式攻擊、對抗式防禦、對抗式訓練、對比式學習、自監督式表示學習、Adversarial attack、Adversarial defense、Adversarial training、Contrastive learning、Self-supervised representation learning

Variational Attention and Disentanglement with Regularization for Conversational Question Answering

電信工程研究所

深度學習、序列學習、序列到序列學習、互信息、潛在解纏、信息理論正則化、變分注意力、貝葉斯神經網絡、聊天機器人、對話式問答、視覺對話、視覺問答、deep learning、sequence learning、sequence to sequence learning、mutual information、latent disentanglement、information-theoretic regularization、variational attention、Bayesian neural network、chatbot、onversational question answering、visual dialog、visual question answering

Contrastive Disentangled Memory for Sequential Learning

電信工程研究所

深度學習、序列對序列學習、對比學習、自監督學習、注意力模型、轉換器、語言模型、自然語言處理、機器翻譯、語音辨識、記憶體取代機制、deep learning、sequence-to-sequence learning、contrastive Learning、self-supervised learning、attention network、transformer、language model、natural language processing、machine translation、speech recognition、memory replacement mechanism

AVAST: Attentive Variational State Tracker for Vision-and-Language Navigation

人工智慧技術與應用碩士學位學程

語言視覺導航、變異推論法、注意力神經網路、部分可觀察馬可夫決策過程、強化學習、經驗回放、vision-and-language navigation、variational inference、attention network、partially observable Markov decision process、reinforcement learning、experience replay

Disentangled Mask Attention in Transformers

電機工程學系

轉換器、序列式學習、注意力機制、深度聚類、解纏式特徵學習、機器翻譯、語音識別、Transformer、sequential learning、attention mechanism、deep clustering、disentangled representation learning、machine translation、speech recognition

Distinctive Augmentation Policies for Natural Language Understanding

電機工程學系

文本分類、資料增強、對比學習、一致性學習、自然語言理解、自然語言處理、text classification、data augmentation、contrastive learning、consistency learning、natural language understanding、natural language processing

Informative Skill Embeddings for Meta Reinforcement Learning

電機工程學系

深度學習、强化學習、元學習、表示學習、資訊瓶頸、連續控制、柔性行動者評論家算法、deep learning、reinforcement learning、meta learning、representation learning、information bottleneck、continuous control、soft actor critic

Adversarial Masking Transformers for Language Understanding

電信工程研究所

深度學習、語言理解、對抗學習、掩蔽模型、機器翻譯、轉換器、deep learning、language understanding、adversarial learning、transformers、masking model、machine translation

Multi-task Generative Adversarial Imitation Learning for Dialogue System

電信工程研究所

強化學習、模仿學習、生成對抗網絡、任務導向對話系統、多任務問題、分層強化學習、稀疏獎勵、reinforcement learning、imitation learning、generative adversarial network、task-oriented dialog system、multi-task problem、hierarchical reinforcement learning、sparse reward

Self-Supervised Learning for Online Speaker Diarization

電機工程學系

即時語者分群、自我監督學習、對比學習、雙向長短時記憶迴圈神經網路、時延神經網路、即時聚類、Online speaker diarization、Self-supervised learning、Contrastive learning、Bi-LSTM、TDNN、Online clustering

Online Compressive Transformer for Speech Recognition

電機工程學系

深度學習、即時語音辨識、壓縮轉感器、連續時間分類、傳感器模型、deep learning、online speech recognition、compressive transformer、connectionist temporal classification、transducer-based model

Model-Based Soft Actor-Critic

電機工程學系

深度學習、強化學習、行動者評論家算法、最大化熵學習、基於模型強化學習、連續空間控制、deep learning、reinforcement learning、soft actor-critic、model-based RL、planning、maximized entropy learning、continuous control、actor-critic

Lower Triangular Flows for Dialog VAE

電機工程學系

後驗塌陷、對話生成、文本生成、歸一化流、變異自動編碼器、序列到序列學習、posterior collapse、dialogue generation、text generation、normalizing flow、variational autoencoder、sequence to sequence learning

Latent Disentanglement in Flow-based Domain Mapping

電機工程學系

生成模型、流模型、潛在解纏結、相互資訊、領域映射、圖像生成、人臉說話圖像生成、generative model、flow-based model、latent disentanglement、mutual information、domain mapping、image synthesis、talking face generation

Graph-Evolved Transformer for Dynamic Network Representation

電機工程學系

深度學習、歸納學習、圖神經網絡、轉換器、動態圖、圖嵌入、金融技術、deep learning、inductive learning、graph neural network、transformer、dynamic graphs、graph embedding、financial technology

DualFormer: An Unified Bidirectional Sequence-to-Sequence Learning

電機工程學系

深度學習、對偶學習、序列對序列學習、注意力模型、轉換器、語言模型、序列對序列模型、自然語言處裡、機器翻譯、deep learning、dual learning、sequence-to-sequence learning、attention network、transformer、language model、sequence-to-sequence model、natural language processing、machine translation

Continuous-Time Attention for Sequential Learning

電機工程學系

深度學習、序列學習、常微分方程、注意力機制、遞迴神經網路、情緒辨識、動作辨識、醫療資料分析、deep learning、sequential learning、ordinary differential equation、attention mechanism、recurrent neural network、emotion recognition、action recognition、medical data analysis

Stochastic Convolutional Recurrent Networks

電機工程學系

深度學習、序列學習、自迴歸生成、變分推理、捲積神經網路、遞迴神經網路、deep learning、sequential learning、autoregressive generation、variational inference、convolutional neural network、recurrent neural network

Adaptive Quantization for Deep Bayesian Neural Network

電信工程研究所

深度學習、網路壓縮、量化神經網路、變分推理、貝氏神經網路、變分自編碼器、Deep learning、quantized neural network、variational inference、Bayesian neural network、variational autoencoder

Stochastic Curiosity Maximizing Exploration

電信工程研究所

深度學習、強化式學習、稀疏獎勵、好奇心、內在獎勵、探索、互消息、deep learning、reinforcement learning、sparse reward、curiosity、intrinsic reward、exploration、mutual information

Neural Variational Inference for Sequence Generation

電機工程學系

後驗塌陷、序列生成、變分自動編碼器、序列到序列學習、歸一化流、posterior collapse、variational autoencoder、sequence to sequence learning、sequence generation、normalizing flow

Stochastic Temporal Difference Learning and Planning

電機工程學系

深度學習、強化式學習、生成模型、序列學習、變分自動編碼器、時間差異、deep learning、reinforcement learning、generative model、sequential learning、variational auto-encoder、temporal difference

Variational Bayesian Generative Adversarial Networks

電機工程學系

深度類神經網路、對抗式學習、i-vector、資料擴增、貝式學習、語者辨識、圖像生成、圖像分類、deep neural network、adversarial learning、Bayesian learning、speaker recognition、image generation、image classification

Deep Hierarchical Sequence Generation with Self-Attention

電信工程研究所

序列生成、變分自動編碼器、序列到序列學習、結構性潛在變數模型、隨機性循環神經網路、注意力機制、sequence generation、variational autoencoder、sequence-to-sequence learning、hierarchical latent variable model、stochastic recurrent neural network、attention mechanism

Hindsight Proximal Policy Optimization

電機資訊國際學程

深度學習、增強式學習、連續控制、策略優化、稀疏獎勵、探索、採樣效率、deep learning、reinforcement learning、continuous control、policy optimization、sparse reward、exploration、sample efficiency

Distributional Policy Optimization for Multi-Goal Reinforcement Learning

電機工程學系

深度學習、增強式學習、策略優化、多目標增強式學習、機器控制、deep learning、reinforcement learning、policy optimization、multi-goal reinforcement learning、robotic control

Neural Meta Learning for Multifidelity Bayesian Optimization

電機工程學系

貝氏最佳化、高斯過程、多細緻度、學會學習法、深層學習、遞歸神經網絡、Bayesian optimization、Gaussian process、multifidelity、meta learning、deep learning、recurrent neural network

Wasserstein Adversarial Domain Adaptation

電機工程學系

深層學習、生成模型、領域調適、隱藏特徵、對抗式學習、部分遷移、最佳化傳輸、deep learning、generative model、domain adaptation、latent features、adversarial learning、partial transfer、optimal transport

Exploring State Transition Uncertainty in Deep Reinforcement Learning

電機工程學系

深度學習、增強式學習、演員-評論家、轉移不確定評論家、狀態轉移、熵搜索、電腦視覺、物件追蹤、物件偵測、deep learning、reinforcement learning、actor-critic、transition uncertainty critic、state transition、entropy search、computer vision、object tracking、object detection

Markov Recurrent Neural Networks for Sequential Deep Learning

電機工程學系

深度學習、序列對序列學習、變分推理、馬可夫鏈、遞迴神經網路、Gumbel-softmax、自然語言處理、語言模型、語音分離、deep learning、sequence to sequence learning、variational inference、Markov chain、recurrent neural network、natural language processing、language model、speech separation

Deep Adversarial Learning for Speaker Recognition

電機工程學系

深度類神經網路、對抗式學習、機率線性鑑別分析、i-vector、資料擴增、流形學習、鑑別式學習、語者辨識、deep neural network、adversarial learning、probabilistic linear discriminant analysis、data augmentation、manifold learning、discriminative learning、speaker recognition

Deep Supporting Attention in Memory Network

電機工程學系

自然語言處理、記憶體加強神經網路、序列對序列學習、注意力機制、問答系統、影像標題生成、natural language processing、memory-augmented neural network、sequence-to-sequence learning、attention mechanism、question answering、image caption

Multiple Target Prediction for Deep Reinforcement Learning

電信工程研究所

深度學習、強化學習、深度Q網路、多重目標預測、自然語言處理、deep learning、reinforcement learning、deep Q network、multiple target prediction、natural language processing

Deep Adversarial Learning and Domain Adaptation

電機工程學系

深層學習、生成模型、潛在特徵、領域調適、對抗式學習、模式識別、deep learning、generative model、domain adaptation、latent features、adversarial learning、pattern classification

Deep Learning for Memory Augmented Speech Separation

電信工程研究所

深度學習、神經圖靈機、序列到序列式學習、記憶體擴增神經網路、訊號分離、語音增強、deep learning、neural Turing machine、sequence to sequence learning、memory-augmented neural network、source separation、speech enhancement

Deep Variational Transfer Learning and Classification

電機工程學系

轉移學習、模式識別、競爭式類神經網路、潛在變數模型、變異性自動編碼器、半監督式學習、分布匹配、transfer learning、pattern recognition、adversarial neural network、latent variable model、variational auto-encoder、semi-supervised learning、distribution matching

Deep Variational Manifold Learning for Speaker Recognition

電機工程學系

流形學習、非線性降維法、深層類神經網路、變異推論、鑑別式學習、語者辨識、manifold learning、nonlinear dimension reduction、deep neural network、variational inference、discriminative learning、speaker recognition

Deep Factorized and Variational Learning for Speech Recognition

電信工程研究所

語音識別、深層學習、矩陣分解、變異型自編碼網路、類神經網路、deep learning、speech recognition、matrix factorization、variational aoto-encoder、neural network

Deep Factorized and Variational Learning for Source Separation

電機工程學系

深層學習、訊號分離、回響消除、類神經網路、矩陣分解、變異型自編碼網路、deep learning、source separation、dereverberation、neural network、matrix factorization、variational auto-encoder

Bayesian Learning for Speech Dereverberation

電信工程研究所

迴響消除、線上學習、貝氏模型、變異性貝氏、非負矩陣分解、speech dereverberation、online learning、Bayesian modeling、variational Bayesian、nonnegative matrix factorization

Stabilization in Optimization for DNN Speech Recognition

電機工程學系

深層學習、隨機梯度下降法、尼氏加速梯度下降法、隨機對偶坐標上升法、最佳化演算法、語音辨識、deep learning、stochastic gradient descent、Nesterov’s accelerated stochastic gradient descent、stochastic dual coordinated ascent、speech recognition、optimization algorithm

Deep Bayesian Inference for Topic Models

電機工程學系

主題模型、貝氏推論、深層學習、狄氏混合分佈、深層展開、Topic models、Bayesian inference、Deep learning、Dirichlet mixture model、deep unfolding

Tensor Neural Networks for Multi-way Data Classification

電信工程研究所

多通路資料、張量分解、類神經網路、深層學習、模式識別、影像辨識、multi-way data、tensor factorization、neural network、deep learning、pattern classification、image recognition

Deep Discriminative Manifold Learning for Speaker Recognition

電機工程學系

流形學習、非線性降維法、深層類神經網路、鑑別式學習、語者辨識、manifold learning、nonlinear dimensionality reduction、deep neural network、discriminative learning、speaker recognition

Deep Semi-Supervised Learning for Domain Adaptation

電信工程研究所

深層學習、領域調適、半監督式學習、轉移學習、deep learning、domain adaptation、semi-supervised learning、transfer learning

Bayesian nonnegative matrix factorization for monaural audio source separation

電信工程研究所

非負矩陣分解、模型選擇、貝氏學習、單通道訊號分離、歌聲分離、nonnegative matrix factorization、model selection、Bayesian learning、monaural source separation、singing voice separation

Tikhonov regularization for deep recurrent neural network acoustic modeling

電信工程研究所

模型正規化、深層學習、遞迴類神經網路、聲學模型、語音辨識、Tikhonov regularization、deep learning、recurrent neural network、acoustic model、speech recognition

Bayesian recurrent neural networks for language modeling

電信工程研究所

貝氏學習、遞迴式類神經網路、海森矩陣、快速近似法、語言模型、語音辨識、Bayesian learning、recurrent neural network、Hessian matrix、rapid approximation、language model、speech recognition