diff --git a/TERMINOLOGY.md b/TERMINOLOGY.md index bad17b497..89f888784 100644 --- a/TERMINOLOGY.md +++ b/TERMINOLOGY.md @@ -2,22 +2,10 @@ 鞍点,saddle point -上下文变量,context variable - -上下文窗口,context window - -上下文词,context word - -上下文向量,context vector - 变换,transform 编码器,encoder -词元,token - -词元化,tokenize - 标签,label 步幅,stride @@ -30,6 +18,8 @@ 层序softmax,hierarchical softmax +查准率,precision + 成本,cost 词表,vocabulary @@ -38,6 +28,12 @@ 词向量,word vector +词元,token + +词元分析器,tokenizer + +词元化,tokenize + 汇聚层,pooling layer 稠密,dense @@ -48,8 +44,6 @@ 轮,epoch -迭代模型参数,update model parameter(s) - 暂退法,dropout 动量法,momentum (method) @@ -64,9 +58,9 @@ 二元分类,binary classification -二元语法,bigram +二元,bigram -二次采样,subsample +子采样,subsample 发散,diverge @@ -76,30 +70,20 @@ 方差,variance -访问参数,access parameters - -非凸优化,non-convex optimization - -分桶,bucketing - 分类,classification 分类器,classifier -分词方式,tokenizer/tokenization - 负采样,negative sampling 感受野,receptive field 格拉姆矩阵,Gram matrix -共现词频,co-occurrence frequency +共现,co-occurrence 归一化,normalization -过滤器,filter - 过拟合,overfitting 核回归,kernel regression @@ -116,8 +100,6 @@ 近似法,approximate method -精确度,precision - 经验风险最小化,empirical risk minimization 局部最小值,local minimum @@ -140,27 +122,25 @@ 拉普拉斯平滑,Laplace smoothing -拼接,concatenate +连结,concatenate 类,class 交叉熵,cross-entropy -连续词袋模型,continous bag-of-words (CBOW) model +连续词袋,continous bag-of-words (CBOW) 零张量,zero tensor 流水线,pipeline -门控循环单元,gated recurrent units (GRU) - -模型参数,model parameter +滤波器,filter -模型复杂度,model complexity +门控循环单元,gated recurrent units (GRU) 目标检测,object detection -偏差,bias +偏置,bias 偏导数,partial derivative @@ -168,23 +148,29 @@ 批量,batch -平均汇聚层,average pooling layer - 齐普夫定律,Zipf's law 欠拟合,underfitting 情感分析,sentiment analysis -全连接层,fully connected layer +全连接层,fully-connected layer 权重,weight -三元语法,trigram +三元,trigram 上采样,upsample -实例,instance +上下文变量,context variable + +上下文窗口,context window + +上下文词,context word + +上下文向量,context vector + +实例/示例,instance 收敛,converge @@ -192,9 +178,11 @@ 数值方法,numerical method -数据集,data set +数据集,dataset -数据样本,data instance +数据示例,data instance + +数据样例,data example 顺序分区,sequential partitioning @@ -220,7 +208,7 @@ softmax回归,softmax regression 填充,padding -跳字模型,skip-gram model +跳元模型,skip-gram model 调参,tune hyperparameter @@ -232,15 +220,15 @@ softmax回归,softmax regression 图像,image -未知词标记,unknown token +未知词元,unknown token 无偏估计,unbiased estimate 误差,error -小批量,mini-batch +小批量,minibatch -小批量梯度,mini-batch gradient +小批量梯度,minibatch gradient 线性模型,linear model @@ -254,11 +242,11 @@ softmax回归,softmax regression 循环神经网络,recurrent neural network (RNN) -样本,example +样例,example 一维梯度下降,gradient descent in one-dimensional space -一元语法,unigram +一元,unigram 隐藏变量,hidden variable @@ -274,8 +262,6 @@ softmax回归,softmax regression 真实值,ground truth -正确率,accuracy - 指标,metric 支持向量机,support vector machine @@ -286,4 +272,4 @@ softmax回归,softmax regression 注意力提示,attention cue -准确率,accuracy +准确率/精度,accuracy