作者:chen_h
微信号 & QQ:862251340
微信公众号:coderpai
简书地址:https://www.jianshu.com/p/b5cd84ac2600
介绍一些人工智能技术的术语,如果你还有术语补充,请访问 Github
| English Terminology | 中文术语 |
|---|---|
| neural networks | 神经网络 |
| activation function | 激活函数 |
| hyperbolic tangent | 双曲正切函数 |
| bias units | 偏置项 |
| activation | 激活值 |
| forward propagation | 前向传播 |
| feedforward neural network | 前馈神经网络 |
| Backpropagation Algorithm | 反向传播算法 |
| (batch) gradient descent | (批量)梯度下降法 |
| (overall) cost function | (整体)代价函数 |
| squared-error | 方差 |
| average sum-of-squares error | 均方差 |
| regularization term | 规则化项 |
| weight decay | 权重衰减 |
| bias terms | 偏置项 |
| Bayesian regularization method | 贝叶斯规则化方法 |
| Gaussian prior | 高斯先验概率 |
| MAP | 极大后验估计 |
| maximum likelihood estimation | 极大似然估计 |
| activation function | 激活函数 |
| tanh function | 双曲正切函数 |
| non-convex function | 非凸函数 |
| hidden (layer) units | 隐藏层单元 |
| symmetry breaking | 对称失效 |
| learning rate | 学习速率 |
| forward pass | 前向传导 |
| hypothesis | 假设值 |
| error term | 残差 |
| weighted average | 加权平均值 |
| feedforward pass | 前馈传导 |
| Hadamard product | 阿达马乘积 |
| forward propagation | 前向传播 |
| off-by-one error | 缺位错误 |
| bias term | 偏置项 |
| numerically checking | 数值检验 |
| numerical roundoff errors | 数值舍入误差 |
| significant digits | 有效数字 |
| unrolling | 组合扩展 |
| learning rate | 学习速率 |
| Hessian matrix Hessian | 矩阵 |
| Newton’s method | 牛顿法 |
| conjugate gradient | 共轭梯度 |
| step-size | 步长值 |
| Autoencoders | 自编码算法 |
| Sparsity | 稀疏性 |
| neural networks | 神经网络 |
| supervised learning | 监督学习 |
| unsupervised learning | 无监督学习 |
| hidden units | 隐藏神经元 |
| the pixel intensity value | 像素灰度值 |
| IID | 独立同分布 |
| PCA | 主元分析 |
| active | 激活 |
| inactive | 抑制 |
| activation function | 激活函数 |
| activation | 激活度 |
| the average activation | 平均活跃度 |
| sparsity parameter | 稀疏性参数 |
| penalty term | 惩罚因子 |
| KL divergence | KL 散度 |
| Bernoulli random variable | 伯努利随机变量 |
| overall cost function | 总体代价函数 |
| backpropagation | 后向传播 |
| forward pass | 前向传播 |
| gradient descent | 梯度下降 |
| the objective | 目标函数 |
| the derivative checking method | 梯度验证方法 |
| Visualizing | 可视化 |
| Autoencoder | 自编码器 |
| hidden unit | 隐藏单元 |
| non-linear feature | 非线性特征 |
| activate | 激励 |
| trivial answer | 平凡解 |
| norm constrained | 范数约束 |
| sparse autoencoder | 稀疏自编码器 |
| norm bounded | 有界范数 |
| input domains | 输入域 |
| vectorization | 矢量化 |
| Logistic Regression | 逻辑回归 |
| batch gradient ascent | 批量梯度上升法 |
| intercept term | 截距 |
| the log likelihood | 对数似然函数 |
| derivative | 导函数 |
| gradient | 梯度 |
| vectorization | 向量化 |
| forward propagation | 正向传播 |
| backpropagation | 反向传播 |
| training examples | 训练样本 |
| activation function | 激活函数 |
| sparse autoencoder | 稀疏自编码网络 |
| sparsity penalty | 稀疏惩罚 |
| average firing rate | 平均激活率 |
| Principal Components Analysis | 主成份分析 |
| whitening | 白化 |
| intensity | 亮度 |
| mean | 平均值 |
| variance | 方差 |
| covariance matrix | 协方差矩阵 |
| basis | 基 |
| magnitude | 幅值 |
| stationarity | 平稳性 |
| normalization | 归一化 |
| eigenvector | 特征向量 |
| redundant | 冗余 |
| variance | 方差 |
| smoothing | 平滑 |
| dimensionality reduction | 降维 |
| regularization | 正则化 |
| reflection matrix | 反射矩阵 |
| decorrelation | 去相关 |
| Principal Components Analysis (PCA) | 主成分分析 |
| zero-mean | 均值为零 |
| mean value | 均值 |
| eigenvalue | 特征值 |
| symmetric positive semi-definite matrix | 对称半正定矩阵 |
| numerically reliable | 数值计算上稳定 |
| sorted in decreasing order | 降序排列 |
| singular value | 奇异值 |
| singular vector | 奇异向量 |
| vectorized implementation | 向量化实现 |
| diagonal | 对角线 |
| Softmax Regression | Softmax回归 |
| supervised learning | 有监督学习 |
| unsupervised learning | 无监督学习 |
| deep learning | 深度学习 |
| logistic regression | logistic回归 |
| intercept term | 截距项 |
| binary classification | 二元分类 |
| class labels | 类型标记 |
| hypothesis | 估值函数/估计值 |
| cost function | 代价函数 |
| multi-class classification | 多元分类 |
| weight decay | 权重衰减 |
| self-taught learning | 自我学习/自学习 |
| unsupervised feature learning | 无监督特征学习 |
| autoencoder | 自编码器 |
| semi-supervised learning | 半监督学习 |
| deep networks | 深层网络 |
| fine-tune | 微调 |
| unsupervised feature learning | 非监督特征学习 |
| pre-training | 预训练 |
| Deep Networks | 深度网络 |
| deep neural networks | 深度神经网络 |
| non-linear transformation | 非线性变换 |
| represent compactly | 简洁地表达 |
| part-whole decompositions | “部分-整体”的分解 |
| parts of objects | 目标的部件 |
| highly non-convex optimization problem | 高度非凸的优化问题 |
| conjugate gradient | 共轭梯度 |
| diffusion of gradients | 梯度的弥散 |
| Greedy layer-wise training | 逐层贪婪训练方法 |
| autoencoder | 自动编码器 |
| Greedy layer-wise training | 逐层贪婪训练法 |
| Stacked autoencoder | 栈式自编码神经网络 |
| Fine-tuning | 微调 |
| Raw inputs | 原始输入 |
| Hierarchical grouping | 层次型分组 |
| Part-whole decomposition | 部分-整体分解 |
| First-order features | 一阶特征 |
| Second-order features | 二阶特征 |
| Higher-order features | 更高阶特征 |
| Linear Decoders | 线性解码器 |
| Sparse Autoencoder | 稀疏自编码 |
| input layer | 输入层 |
| hidden layer | 隐含层 |
| output layer | 输出层 |
| neuron | 神经元 |
| robust | 鲁棒 |
| sigmoid activation function | S型激励函数 |
| tanh function | tanh激励函数 |
| linear activation function | 线性激励函数 |
| identity activation function | 恒等激励函数 |
| hidden unit | 隐单元 |
| weight | 权重 |
| error term | 偏差项 |
| Full Connected Networks | 全连接网络 |
| Sparse Autoencoder | 稀疏编码 |
| Feedforward | 前向输送 |
| Backpropagation | 反向传播 |
| Locally Connected Networks | 部分联通网络 |
| Contiguous Groups | 连接区域 |
| Visual Cortex | 视觉皮层 |
| Convolution | 卷积 |
| Stationary | 固有特征 |
| Pool | 池化 |
| features | 特征 |
| example | 样例 |
| over-fitting | 过拟合 |
| translation invariant | 平移不变性 |
| pooling | 池化 |
| extract | 提取 |
| object detection | 物体检测 |
| DC component | 直流分量 |
| local mean subtraction | 局部均值消减 |
| sparse autoencoder | 消减归一化 |
| rescaling | 缩放 |
| per-example mean subtraction | 逐样本均值消减 |
| feature standardization | 特征标准化 |
| stationary | 平稳 |
| zero-mean | 零均值化 |
| low-pass filtering | 低通滤波 |
| reconstruction based models | 基于重构的模型 |
| RBMs | 受限Boltzman机 |
| k-Means | k-均值 |
| long tail | 长尾 |
| loss function | 损失函数 |
| orthogonalization | 正交化 |
| Sparse Coding | 稀疏编码 |
| unsupervised method | 无监督学习 |
| over-complete bases | 超完备基 |
| degeneracy | 退化 |
| reconstruction term | 重构项 |
| sparsity penalty | 稀疏惩罚项 |
| norm | 范式 |
| generative model | 生成模型 |
| linear superposition | 线性叠加 |
| additive noise | 加性噪声 |
| basis feature vectors | 特征基向量 |
| the empirical distribution | 经验分布函数 |
| the log-likelihood | 对数似然函数 |
| Gaussian white noise | 高斯白噪音 |
| the prior distribution | 先验分布 |
| prior probability | 先验概率 |
| source features | 源特征 |
| the energy function | 能量函数 |
| regularized | 正则化 |
| least squares | 最小二乘法 |
| convex optimization software | 凸优化软件 |
| conjugate gradient methods | 共轭梯度法 |
| quadratic constraints | 二次约束 |
| the Lagrange dual | 拉格朗日对偶函数 |
| feedforward architectures | 前馈结构算法 |
| Independent Component Analysis | 独立成分分析 |
| Over-complete basis | 超完备基 |
| Orthonormal basis | 标准正交基 |
| Sparsity penalty | 稀疏惩罚项 |
| Under-complete basis | 不完备基 |
| Line-search algorithm | 线搜索算法 |
| Topographic cost term | 拓扑代价项 |
作者:chen_h
微信号 & QQ:862251340
简书地址:https://www.jianshu.com/p/b5cd84ac2600
CoderPai 是一个专注于算法实战的平台,从基础的算法到人工智能算法都有设计。如果你对算法实战感兴趣,请快快关注我们吧。加入AI实战微信群,AI实战QQ群,ACM算法微信群,ACM算法QQ群。长按或者扫描如下二维码,关注 “CoderPai” 微信号(coderpai)
AI术语解析
2229

被折叠的 条评论
为什么被折叠?



