人工智能詞匯12.17文件【新版】

1、精常用英語詞匯-andrew Ng 課程training set 訓練集合inten sity 強度Regressi on回歸Loss fun ction損失函數(shù)non-convex 勻非凸函數(shù)neural network神經(jīng)網(wǎng)絡supervised lear ning監(jiān)督學習regressi on problem回歸問題處理的是連續(xù)的問題classificati on problem分類問題discreet value 離散值support vector mach ines 支持向量機 learning theory 學習理論learning algorithms學習算法un supervi

2、sed lear ning無監(jiān)督學習gradie nt desce nt梯度下降lin ear regressi on線性回歸Neural Network神經(jīng)網(wǎng)絡gradie nt desce nt梯度下降no rmal equatio nslin ear algebra線性代數(shù)superscript 上標exp onen tiati on扌旨數(shù)trai ning example訓練樣本hypothesis 假設，用來表示學習算法的輸出LMS algorithmleast mean squares 最小二乘法算法batch gradie nt desce nt批量梯度下降con sta ntl

3、y gradie nt desce nt隨機梯度下降iterative algorithm 迭代算法partial derivative 偏導數(shù)con tour等高線quadratic fun cti on二元函數(shù)locally weighted regressi on局部加權(quán)回歸un derfitt ing欠擬合overfitt ing過擬合non-parametriclear ningalgorithms無參數(shù)學習算法parametric lear ning algorithm參數(shù)學習算法activatio n 激活值activati on fun cti on激活函數(shù)additive n

4、oise加性噪聲autoe ncoder自編碼器Autoe ncoders自編碼算法average firing rate平均激活率精精average sum-of-squares error均方差diago nal對角線backpropagati on后向傳播basis 基basis feature vectors特征基向量batch gradie nt asce nt批量梯度上升法Bayesia nregularizati onmethod貝葉斯規(guī)則化方法Berno ulli ran dom variable伯努利隨機變量bias term 偏置項binary classficati on

5、二元分類class labels類型標記con cate nati on級聯(lián)diffusion of gradie nts梯度的彌散eige nvalue 特征值eige nvector特征向量error term 殘差feature matrix特征矩陣feature sta ndardizatio n特征標準化feedforward architectures前饋結(jié)構(gòu)算法feedforward n eural n etwork前饋神經(jīng)網(wǎng)絡feedforward pass前饋傳導fin e-tu ned 微調(diào)first-order feature一階特征精conjugate gradie n

6、t共軛梯度con tiguous groups聯(lián)通區(qū)域convex optimizati on software凸優(yōu)化軟件con voluti on卷積cost function代價函數(shù)covaria nee matrix協(xié)方差矩陣DC comp on e nt直流分量decorrelati on 去相關dege neracy退化deme nsion ality reductio n降維forward pass 前向傳導forward propagati on前向傳播Gaussia n prior咼斯先驗概率gen erative model生成模型gradie nt desce nt梯度下降

7、Greedy layer-wise trai ning逐層貪婪訓練方法group ing matrix分組矩陣Hadamard product阿達馬乘積Hessia n matrix Hessia n 矩陣hidden layer 隱含層derivative 導函數(shù)精hidden units隱藏神經(jīng)元lin ear superpositi on線性疊力口Hierarchical group ing層次型分組higher-order features更高階特征highly non-convex optimizati on problem高度非凸的優(yōu)化問題histogram 直方圖hyperbol

8、ic tan ge nt雙曲正切函數(shù)hypothesis 估值，假設ide ntity activation fun ctio n恒等激勵函數(shù)IID獨立同分布illu min ati on照明in active 抑制in depe ndent comp onent an alysis獨立成份分析in put doma ins輸入域input layer 輸入層in te nsity亮度/灰度in tercept term截距KL diverge nee相對熵KL diverge nee KL 分散度k-Mea ns K-均值learning rate學習速率least squares 最小二乘

9、法lin e-search algorithm線搜索算法local mean subtract ion局部均值消減local optima 局部最優(yōu)解logistic regressi on邏輯回歸loss fun ctio n損失函數(shù)low-pass filteri ng低通濾波magn itude 幅值MAP極大后驗估計maximum likelihood estimati on極大似然估計mea n 平均值MFCC Mel倒頻系數(shù)multi-class classificati on多元分類neural networks神經(jīng)網(wǎng)絡neuron 神經(jīng)元Newton s method 牛頓法n

10、on-con vex fun ctio n耳非凸函數(shù)non-I in ear feature 非線性特征n orm 范式norm boun ded有界范數(shù)norm constrained范數(shù)約束no rmalizati on 歸一化精lin ear corresp ondence線性響應精nu merical roun doff errorsrecon struct ion cost重建代價output layer 輸出層overall cost fun ctio n總體代價函數(shù)over-complete basis超完備基over-fitt ing過擬合parts of objects目標的

11、部件part-whole decomposti on部分-整體分解PCA主元分析pen alty term懲罰因子per-example mean subtracti on逐樣本均值消減pooli ng 池化pretrain 預訓練prin cipal comp onents an alysis主成份分析quadratic con stra ints二次約束RBMs 受限 Boltzman 機recon struct ion based models基于重構(gòu)的模型recon struct ion term重構(gòu)項redundant 冗余reflection matrix反射矩陣regulariz

12、ati on 正則化regularizati on term正則化項rescali ng 縮放robust 魯棒性run 行程sec on d-order feature二階特征sigmoid activation fun ctio n S型激勵函數(shù)sig ni fica nt digits有效數(shù)字sin gular value 奇異值sin gular vector奇異向量smoothed L1 penalty平滑的L1范數(shù)懲罰Smoothed topographic L1 sparsity pen alty平滑地形L1稀疏懲罰函數(shù)smooth ing平滑Softmax Regress on

13、 Softmax 回歸sorted in decreas ing order降序排歹 Unu merically check ing數(shù)值檢驗nu merically reliable數(shù)值計算上穩(wěn)定object detect ion物體檢測objective function目標函數(shù)off-by-one error缺位錯誤orthogo nalizati on正交化數(shù)值舍入誤差精source features 源特征sparse autoe ncoder消減歸一化精Sparsity 稀疏性tran sformati on變換sparsity parameter稀疏性參數(shù)sparsity pen

14、alty稀疏懲罰square function平方函數(shù)squared-error 方差statio nary平穩(wěn)性（不變性）stati onary stochastic process平穩(wěn)隨機過程step-size 步長值supervised lear ning監(jiān)督學習symmetric positive semi-defi nite matrix對稱半正定矩陣symmetry break ing對稱失效tanh fun ction雙曲正切函數(shù)the average activati on平均活躍度the derivative check ing method梯度驗證方法the empiric

15、al distributio n經(jīng)驗分布函數(shù)the en ergy function能量函數(shù)the Lagra nge dual拉格朗日對偶函數(shù)the log likelihood對數(shù)似然函數(shù)the pixel inten sity value像素灰度值the rate of con verge nee收斂速度topographic cost term拓撲代價項topographic ordered拓撲秩序tran slati on in varia nt平移不變性trivial an swer 平凡解un der-complete basisun rolli ng 組合擴展un superv

16、ised lear ningvaria nee 方差vecotrized impleme ntati onvectorizati on矢量化visual cortex視覺皮層weight decay權(quán)重衰減weighted average力口權(quán)平均值white ning白化zero-mea n均值為零Accumulated error backpropagati on累積誤差逆?zhèn)鞑ctivatio n Fun ction激活函數(shù)Adaptive Reso nance Theory/ART自適應諧振理論Addictive model 加性學習Adversarial Networks對抗網(wǎng)絡Af

17、fine Layer 仿射層Affinity matrix 親和矩陣Age nt代理/智能體不完備基無監(jiān)督學習向量化實現(xiàn)精精Class-c on diti onal probability類條件概率Alpha-beta pruning a- B剪枝Ano maly detecti on異常檢測Approximati on 近似Area Under ROC Curve / AUC Roc 曲線下面積Artificial Ge neral In tellige nce/AGI通用人工智能Artificial In tellige nce/AI 人工智能Associati on an alysis關

18、聯(lián)分析Atte ntio n mecha nism注意力機制Attribute con diti onal in depe ndence assumptio n屬性條件獨立性假設Attribute space屬性空間Attribute value屬性值Autoe ncoder自編碼器Automatic speech recog niti on自動語音識別Automatic summarizati on自動摘要Average gradie nt平均梯度Average-Pooli ng平均池化Backpropagati onThroughTime通過時間的反向傳播Backpropagatio n/

19、BP反向傳播Base lear ner基學習器Base lear ning algorithm 基學習算法Bayes decisi on rule貝葉斯判定準貝UBayes Model Averaging/ BMA 貝葉斯模型平均Bayes optimal classifier貝葉斯最優(yōu)分類器Bayesia n decisi on theory貝葉斯決策論Bayesian network貝葉斯網(wǎng)絡Betwee n-class scatter matrix類間散度矩陣Bias偏置/偏差Bias-varia nee decompositi on偏差-方差分解Bias-Varianee Dilemm

20、a 偏差 -方差困境Bi-directio nal Lon g-Short Term Memory/Bi-LSTM雙向長短期記憶Binary classificatio n二分類Bi no mial test二項檢驗Bi-partitio n二分法Boltzma nn mach ine玻爾茲曼機Bootstrap sampli ng自助采樣法/可重復采樣Bootstrapp ing 自助法Break-Eve nt Poi nt / BEP 平衡點Calibratio n校準Cascade-Correlation級聯(lián)相關Algorithm 算法Batch Normalizatio n/BN批量歸

21、一化精Categorical attribute離散屬性精Classificati on and regressi on tree/CART分類與回Con ditio nal Probability Table/CPT條件概率表歸樹Classifier 分類器Class-imbala nee類別不平衡Closed -form 閉式Cluster簇/類/集群Cluster an alysis聚類分析Clusteri ng聚類Clusteri ng en semble聚類集成Co-adapti ng 共適應Codi ng matrix編碼矩陣COLT國際學習理論會議Committee-based

22、lear ning基于委員會的學習Competitive lear ning競爭型學習Comp onent lear ner組件學習器Comprehe nsibility可解釋性Computati on Cost計算成本Computati onal Lin guistics計算語言學Computer visio n計算機視覺Con cept drift概念漂移Con cept Learni ng System /CLS 概念學習系統(tǒng)Con ditio nal en tropy條件熵Con ditio nal ran dom field/CRF條件隨機場Con ditio nal risk條件風

23、險Con fide nee 置信度Con fusi on matrix混淆矩陣Connection weight 連接權(quán)Connectionism 連結(jié)主義Con siste ncy 致性/相合性Contingency table 歹 U聯(lián)表Contin uous attribute連續(xù)屬性Con verge nee收斂Con versati onal age nt會話智能體Convex quadratic program ming凸二次規(guī)戈 UCon vexity 凸性Con volutio nal neural network/CNN卷積神經(jīng)網(wǎng)絡Co-occurre nee同現(xiàn)Correl

24、atio n coefficie nt相關系數(shù)Cosi ne similarity余弦相似度Cost curve成本曲線Cost Fun ction成本函數(shù)Cost matrix 成本矩陣精Con diti onal mutual in formati on條件互信息Cost-se nsitive成本敏感Cross en tropy交叉熵Cross validati on交叉驗證Crowdsourc ing 眾包Curse of dime nsi on ality維數(shù)災難Cut poi nt截斷點Cutti ng pla ne algorithm割平面法Data mi ning數(shù)據(jù)挖掘Data

25、 set 數(shù)據(jù)集Decisi on Boundary 決策邊界Decisi on stump 決策樹樁Decisi on tree 決策樹/判定樹Deduction 演繹Deep Belief Network深度信念網(wǎng)絡Deep Con voluti onalGen erativeAdversarialNetworkDCGAN深度卷積生成對抗網(wǎng)絡Deep lear ning深度學習Deep neural network/DNN深度神經(jīng)網(wǎng)絡Deep Q-Learning深度 Q 學習Deep Q-Network深度 Q 網(wǎng)絡Den sity estimati on密度估計Den sity-bas

26、ed clusteri ng密度聚類Differe ntiable neural computer可微分神經(jīng)計算機Dime nsi on ality reduction algorithm降維算法Directed edge 有向邊Disagreeme nt measure不合度量Discrim in ative model判另 U模型Discrim in ator判另 U器Dista nee measure距離度量Dista nee metric lear ning距離度量學習Distributio n分布Diverge nee散度Diversity measure多樣性度量/差異性度量Dom

27、a in adapti on領域自適應Down sampli ng下采樣D-separati on(Directed separatio n) 有向分離Dual problem對偶問題Dummy node啞結(jié)點Dyn amic Fusi on動態(tài)融合Dyn amic program ming動態(tài)規(guī)戈 UEige nvalue decompositi on特征值分解精Embeddi ng嵌入Emoti onal an alysis情緒分析Empirical con diti onal en tropy經(jīng)驗條件熵精Empirical en tropy經(jīng)驗熵Feature select ion特征選

28、擇Empirical error 經(jīng)驗誤差Empirical risk 經(jīng)驗風險En d-to-E nd端至 U端En ergy-based model基于能量的模型En semble lear ning集成學習En semble pruning集成修剪Error CorrectingOutput Codes / ECOC 糾錯輸出碼Error rate 錯誤率Error-ambiguity decomposition誤差-分歧分解Euclidea n dista nee歐氏距離Evoluti onary computatio n演化計算Expectati on-Maximizati on期望

29、最大化Expected loss期望損失Explod ing Gradie nt Problem梯度爆炸問題Exponen tial loss function指數(shù)損失函數(shù)Extreme Learni ng Machi ne/ELM超限學習機Factorizati on因子分解False n egative 假負類False positive 假正類False Positive Rate/FPR假正例率Feature vector特征向量Featured Lear ning特征學習Feedforward Neural Networks/FNN前饋神經(jīng)網(wǎng)絡Fin e-t uning 微調(diào)Flip

30、p ing output 翻轉(zhuǎn)法Fluctuati on震蕩Forward stagewise algorithm前向分步算法Freque ntist 頻率主義學派Full-ra nk matrix滿秩矩陣Functional neuron 功能神經(jīng)元Gain ratio 增益率Game theory 博弈論Gaussia n kernel fun cti on高斯核函數(shù)Gaussian Mixture Model高斯混合模型Gen eral Problem So Iving通用問題求解Gen eralizati on 泛化Gen eralizati on error泛化誤差Gen erali

31、zati on error bound泛化誤差上界Gen eralized Lagra nge function廣義拉格朗日函數(shù)Gen eralized linear model 廣義線性模型Gen eralized Rayleigh quotie nt廣義瑞禾 U商精Feature engin eeri ng特征工程精Gen erativeAdversarial Networks/GAN生成對抗Hold-out 留出法網(wǎng)絡Gen erative Model生成模型Gen erator生成器Gen etic Algorithm/GA遺傳算法Gibbs sampl ing吉布斯采樣Gini in

32、 dex基尼指數(shù)Global mi nimum全局最小Global Optimizati on全局優(yōu)化Gradie nt boost ing梯度提升Gradie nt Desce nt梯度下降Graph theory圖論Grou nd-truth真相/真實Hard margin 硬間隔Hard voti ng 硬投票Harmo nic mea n調(diào)和平均Hesse matrix海塞矩陣Hidde n dyn amic model隱動態(tài)模型Hidden layer 隱藏層Hidden Markov Model/HMM隱馬爾可夫模型Hierarchical clusteri ng層次聚類Hilbe

33、rt space 希爾伯特空間Homoge neous同質(zhì)Hybrid comput ing混合計算Hyperparameter 超參數(shù)Hypothesis 假設Hypothesis test假設驗證ICML國際機器學習會議Improved iterative scali ng/IIS改進的迭代尺度法In creme ntal lear ning增量學習In depe ndent and ide ntically distributed/i.i.d.獨立同分布In depe nde ntCompo nentAn alysis/ICA獨立成分分析In dicator fun cti on指示函數(shù)

34、In dividual learner個體學習器In duction歸納In ductive bias歸納偏好In ductive lear ning歸納學習InductiveLogic Programming / ILP 歸納邏輯程序設計In formati on en tropy信息熵In formatio n gain信息增益Input layer 輸入層Hi nge loss fun ction合頁損失函數(shù)Insen sitive loss不敏感損失精精Intern ati onalConferencefor Mach ine拉格朗日乘子拉普拉斯平滑拉普拉斯修正Log-li near

35、regressi on對數(shù)線性回歸In ter-cluster similarity簇間相似度Learni ng/ICML國際機器學習大會In tra-cluster similarity簇內(nèi)相似度Intrin sic value固有值Isometric Mapp in g/Isomap等度量映射Isot onic regressi on等分回歸Iterative Dichotomiser 迭代二分器Kernel method 核方法Kernel trick 核技巧Kern elized Lin ear Discrim inant An alysis/ KLDA核線性判別分析K-fold cr

36、oss validation k折交叉驗證/ k倍交叉驗證K-Means Clustering K -均值聚類K-Nearest Neighbours Algorithm/KNN K近鄰算法Kno wledge base知識庫Kno wledge Represe ntati on知識表征Label space標記空間Lagrange duality拉格朗日對偶性Lagra nge multiplierLaplace smoothi ngLaplacia n correct ionLate nt Dirichlet Allocatio n隱狄利克雷分布Late nt sema ntic an a

37、lysis潛在語義分析Late nt variable隱變量Lazy learni ng懶惰學習Lear ner 學習器Lear ning by an alogy類比學習Lear ning rate學習率Learni ng Vector Qua ntizati on/LVQ學習向量量化Least squares regressi on tree最小二乘回歸樹Leave-O ne-Out/LOO 留一法lin ear cha in con diti onal ran dom field線性鏈條件隨機場Linear Discriminant Analysis/ LDA 線性判別分析Linear m

38、odel 線性模型Lin ear Regressi on線性回歸Link fun ction聯(lián)系函數(shù)Local Markov property局部馬爾可夫性Local mi nimum局部最小Log likelihood對數(shù)似然Log odds / logit 對數(shù)幾率精Logistic Regressi on Logistic回歸Log-likelihood對數(shù)似然精Lon g-Short Term Memory/LSTM長短期記憶Metric lear ning度量學習Loss fun ctio n損失函數(shù)Machi ne tran slatio n/MT機器翻譯Macro n-P 宏查準

39、率Macro n-R 宏查全率Majority voting 絕對多數(shù)投票法Manifold assumpti on流形假設Man ifold learni ng 流形學習Margin theory 間隔理論Margi nal distributio n邊際分布Marginal in depe ndence邊際獨立性Margin alizati on邊際化Markov Cha in Monte Carlo/MCMC馬爾可夫鏈蒙特卡羅方法Markov Ra ndom Field 馬爾可夫隨機場Maximal clique 最大團Maximum Likelihood Estimatio n/MLE

40、極大似然估計/極大似然法Maximum margin最大間隔Maximum weighted spanning tree 最大帶權(quán)生成 樹Max-Pooli ng最大池化Mean squared error 均方誤差Micro-P微查準率Micro-R 微查全率Mi ni mal Description Len gth/MDL最小描述長度Min imax game極小極大博弈Misclassificatio n cost 誤分類成本Mixture of experts 混合專家Mome ntum 動量Moral graph 道德圖/端正圖Multi-class classificatio n多

41、分類Multi-docume nt summarizati on多文檔摘要Multi-layer feedforward n eural n etworks多層前饋神經(jīng)網(wǎng)絡Multilayer Perceptro n/MLP多層感知器Multimodal learni ng多模態(tài)學習Multiple Dime nsio nal Scali ng多維縮放Multiple lin ear regressio n多元線性回歸Multi-resp onse Lin ear Regressi on/ MLR多響應線性回歸Mutual in formati on互信息Naive bayes樸素貝葉斯Nai

42、ve Bayes Classifier 樸素貝葉斯分類器Named en tity recog ni tio n命名實體識別Meta-learner元學習器精Normalizati on 歸一化Nuclear norm核范數(shù)Numerical attribute數(shù)值屬性Letter OObjective function目標函數(shù)Oblique decisi on tree斜決策樹Occam s razor 奧卡姆剃刀Odds幾率Off-Policy 離策略One shot learni ng次性學習One-Dependent Estimator/ ODE 獨依賴估計On-Policy 在策略O

43、rdi nal attribute有序?qū)傩設ut-of-bag estimate 包外估計Output layer 輸出層Output smeari ng 輸出調(diào)制法Overfitt ing過擬合/過配Oversampli ng過采樣Paired t-test 成對 t 檢驗Pairwise 成對型Pairwise Markov property成對馬爾可夫性Parameter 參數(shù)Nash equilibrium納什均衡Natural la nguage gen eratio n/NLG自然語言生成Natural la nguage process ing自然語言處理Negative cla

44、ss 負類Negative correlati on 負相關法Negative Log Likelihood負對數(shù)似然Neighbourhood Comp onent An alysis/NCA近鄰成分分析Neural Machi ne Tran slation神經(jīng)機器翻譯Neural Turing Machi ne神經(jīng)圖靈機Newton method 牛頓法NIPS國際神經(jīng)信息處理系統(tǒng)會議No Free Lunch Theorem /NFL沒有免費的午餐定理Noise-c on trastive estimati on噪音對比估計Nomi nal attribute列名屬性Non-conve

45、x optimizati on非凸優(yōu)化No nlin ear model非線性模型Non-metric dista nee非度量距離Non-n egative matrix factorizati on非負矩陣分解Non-ordinal attribute無序?qū)傩訬on-Saturati ng Game非飽和博弈精Norm 范數(shù)精Proximal Gradie nt Desce nt/PGD近端梯度下降Pruning 剪枝Pseudo-label 偽標記Qua ntized Neural Network量子化神經(jīng)網(wǎng)絡Quan tum computer量子計算機Quan tum Comput i

46、ng量子計算Quasi Newt on method擬牛頓法Radial Basis Function/RBF徑向基函數(shù)Ran dom Forest Algorithm隨機森林算法Ran dom walk隨機漫步Recall查全率/召回率Receiver Operat ing Characteristic/ROCRectified Lin ear Un it/ReLU線性修正單元Recurre nt Neural Network循環(huán)神經(jīng)網(wǎng)絡Recursive n eural n etwork遞歸神經(jīng)網(wǎng)絡Reference model參考模型Regressi on回歸Regularizati o

47、n正則化Re in forceme nt lear nin g/RL強化學習Represe ntati on lear ning表征學習Prior 先驗Parameter tuning 調(diào)參Parse tree 解析樹Particle Swarm Optimizatio n/PSO粒子群優(yōu)化算法Part-of-speech taggi ng詞性標注Perceptro n感知機Performa nee measure性能度量Plug and Play Generative Network 即插即用生成網(wǎng)絡Plurality voti ng 相對多數(shù)投票法Polarity detectio n 極

48、性檢測Poly nomial kernel fun ctio n多項式核函數(shù)Pooli ng池化Positive class 正類Positive defi nite matrix正定矩陣Post-hoc test后續(xù)檢驗Post-pr uning后剪枝pote ntial function勢函數(shù)Precisi on 查準率/準確率Prepruning 預剪枝Prin cipal comp onent an alysis/PCA主成分分析Prin ciple of multiple expla nati ons多釋原貝 UParameter estimati on參數(shù)估計Probability

49、 Graphical Model概率圖模型受試者工作特征精Represe nter theorem表示定理精reproduci ng kernel Hilbert space/RKHS再生核希爾伯特空間Re-sampli ng重采樣法Rescali ng 再縮放Residual Mappi ng殘差映射Residual Network殘差網(wǎng)絡RestrictedBoltzma nnMachi ne/RBM受限玻爾茲曼機Restricted Isometry Property/RIP限定等距性Re-weighti ng重賦權(quán)法Robustness穩(wěn)健性/魯棒性Root node 根結(jié)點Rule

50、Engine 規(guī)則引擎Rule learni ng規(guī)則學習Saddle poi nt鞍點Sample space樣本空間Sampli ng 采樣Score fu nction評分函數(shù)Self-Drivi ng 自動駕駛Self-Orga niz ing Map/ SOM 自組織映射Semi-naive Bayes classifiers半樸素貝葉斯分類器Semi-Supervised Lear ning半監(jiān)督學習semi-Supervised Support Vector Mach ine半監(jiān)督支持向量機Sen time nt an alysis情感分析Separati ng hyperpla

51、 ne分離超平面Sigmoid fun cti on Sigmoid函數(shù)Similarity measure相似度度量Simulated ann eali ng 模擬退火Simulta neous localizati on and mapp ing同步定位與地圖構(gòu)建Sin gular Value Decompositi on奇異值分解Slack variables松弛變量Smoothi ng 平滑Soft margin 軟間隔Soft margin maximizati on軟間隔最大化Soft voti ng 軟投票Sparse represe ntati on稀疏表征Sparsity 稀

52、疏性Specializati on 特化Spectral Clusteri ng 譜聚類Speech Recog niti on 語音識另 USplitti ng variable 切分變量Squash ing function擠壓函數(shù)Stability-plasticity dilemma 可塑性-穩(wěn)定性困境Statistical learn ing統(tǒng)計學習精Status feature fun ctio n狀態(tài)特征函Time Step 時間步驟Stochastic gradie nt desce nt隨機梯度下降Stratified sampli ng分層采樣Structural risk

53、結(jié)構(gòu)風險Structural risk mi nimizatio n/SRM結(jié)構(gòu)風險最小化Subspace 子空間Supervised learni ng監(jiān)督學習/有導師學習support vector expa nsion支持向量展式Support Vector Mach in e/SVM支持向量機Surrogat loss替代損失Surrogate fu nctio n替代函數(shù)Symbolic lear ning符號學習Symbolism符號主義Syn set同義詞集T-Distributi on Stochastic Neighbour Embedd ingt-SNE T -分布隨機近鄰

54、嵌入Ten sor 張量Ten sor Processi ng Uni ts/TPU張量處理單元The least square method最小二乘法Threshold 閾值Threshold logic un it閾值邏輯單元Threshold-movi ng 閾值移動Toke ni zati on標記化Trai ning error訓練誤差Training in sta nee訓練示例/訓練例Tran sductive lear ning直推學習Transfer lear ning遷移學習Treeba nk樹庫Tria-by-error 試錯法True n egative真負類True

55、positive 真正類True Positive Rate/TPR 真正例率Turi ng Machi ne圖靈機Twice-learni ng二次學習Un derfitt ing 欠擬合/欠配Un dersampli ng欠采樣Un dersta ndability可理解性Un equal cost非均等代價Uni t-step fun ctio n單位階躍函數(shù)Uni variate decisi on tree 單變量決策樹Un supervised lear ning無監(jiān)督學習/無導師學習Un supervised layer-wise trai ning無監(jiān)督逐層訓練Upsampli

56、 ng上采樣精Van ishi ng Gradie nt Problem梯度消失問題axiom公理，原則Wasserstein GAN/WGAN Wasserste in生成對抗網(wǎng)Variati onal inference變分推斷VC Theory VC 維理論Version space版本空間Viterbi algorithm維特比算法Von Neuma nn architecture馮 諾伊曼架構(gòu)絡Weak learner 弱學習器Weight 權(quán)重Weight shari ng 權(quán)共享Weighted voti ng加權(quán)投票法Withi n-class scatter matrix類內(nèi)

57、散度矩陣Word embeddi ng詞嵌入Word sense disambiguati on 詞義消歧Zero-data lear ning零數(shù)據(jù)學習Zero-shot lear ning零次學習approximati ons 近似值arbitrary 隨意的affine仿射的arbitrary 任意的amino acid 氨基酸abstract 提取architecture 架構(gòu)，體系結(jié)構(gòu)；建造業(yè)absolute 絕對的arsenal軍火庫assig nment 分酉己algebra線性代數(shù)asymptotically無癥狀的appropriate 恰當?shù)腷ias偏差brevity簡短，

58、簡潔；短暫800 broader 廣泛briefly簡短的batch批量con verge nee 收斂，集中至U點convex 凸的con tours 輪廓con stra int 約束con sta nt 常理commercial 商務的compleme ntarity 補充coord in ate asce nt同等級上升精ame nable 經(jīng)得起檢驗的精clippi ng 剪下物；剪報；修剪comp onent 分量；部件con ti nu ous連續(xù)的covaria nee 協(xié)方差canoni cal 正規(guī)的，正則的con cave 非凸的correspo nds相符合；相當；通信

59、corollary 推論con crete具體的事物，實在的東西cross validati on交叉驗證correlatio n相互關系conven ti on 約定cluster 一簇cen troids質(zhì)心，形心con verge 收斂computati on ally計算（機）的calculus 計算derive獲得，取得dual 二元的duality 二元性；二象性；對偶性derivatio n 求導；得到；起源denote 預示，表示，是 的標志；意味著,diverge nee散度；發(fā)散性dime nsion 尺度，規(guī)格；維數(shù) dot小圓點distortio n變形density概率密度函數(shù)discrete 離散的discrim in ative 有識別能力的 diago nal 對角dispers ion 分散，散開 determ inant決定因素disjoint不相交的 encoun ter 遇至 Uellipses 橢圓equality 等式extra額外的empirical 經(jīng)驗；觀察ennm erate例舉，計數(shù)exceed 超過，越出 expectatio n 期望 efficie nt 生效的e