k-Means K-均值 

learning rate 学习速率 

least squares 最小二乘法 

linear correspondence 线性响应 

linear superposition 线性叠加 

line-search algorithm 线搜索算法 

local mean subtraction 局部均值消减 

local optima 局部最优解 

logistic regression 逻辑回归 

loss function 损失函数 

low-pass filtering 低通滤波 

magnitude 幅值 

MAP 极大后验估计 

maximum likelihood estimation 极大似然估计 

mean 平均值 

MFCC Mel 倒频系数 

multi-class classification 多元分类 

neural networks 神经网络 

neuron 神经元 

Newton’s method 牛顿法 

non-convex function 非凸函数 

non-linear feature 非线性特征 

norm 范式 

norm bounded 有界范数 

norm constrained 范数约束 

normalization 归一化 

numerical roundoff errors 数值舍入误差 

numerically checking 数值检验 

numerically reliable 数值计算上稳定 

object detection 物体检测 

objective function 目标函数 

off-by-one error 缺位错误 

orthogonalization 正交化 

output layer 输出层 

overall cost function 总体代价函数 

over-complete basis 超完备基 

over-fitting 过拟合 

parts of objects 目标的部件 

part-whole decompostion 部分-整体分解 

PCA 主元分析 

penalty term 惩罚因子 

per-example mean subtraction 逐样本均值消减 

pooling 池化 

pretrain 预训练 

principal components analysis 主成份分析 

quadratic constraints 二次约束 

RBMs 受限Boltzman机 

reconstruction based models 基于重构的模型 

reconstruction cost 重建代价 

reconstruction term 重构项 

redundant 冗余 

reflection matrix 反射矩阵 

regularization 正则化 

regularization term 正则化项 

rescaling 缩放 

robust 鲁棒性 

run 行程 

second-order feature 二阶特征 

sigmoid activation function S型激励函数 

significant digits 有效数字 

singular value 奇异值 

singular vector 奇异向量 

smoothed L1 penalty 平滑的L1范数惩罚 

Smoothed topographic L1 sparsity penalty 平滑地形L1稀疏惩罚函数 

smoothing 平滑 

Softmax Regresson Softmax回归 

sorted in decreasing order 降序排列 

source features 源特征 

sparse autoencoder 消减归一化 

Sparsity 稀疏性 

sparsity parameter 稀疏性参数 

sparsity penalty 稀疏惩罚 

square function 平方函数 

squared-error 方差 

stationary 平稳性（不变性） 

stationary stochastic process 平稳随机过程 

step-size 步长值 

supervised learning 监督学习 

symmetric positive semi-definite matrix 对称半正定矩阵 

symmetry breaking 对称失效 

tanh function 双曲正切函数 

the average activation 平均活跃度