【小知识】【含MATLAB源码】利用神经网络降低数据维度

通过训练具有小型中心层的多层神经网络来重构高维输入向量，可以将高维数据转换为低维编码。

High-dimensional data can be converted tolow-dimensional codes by training a multilayer neural network with a smallcentral layer to reconstruct high-dimensional input vectors.

在本文的“自编码”网络中可以使用梯度下降来精准地调节权值，但前提是初始权值必须足够靠近优值解。

Gradient descent can be used forfine-tuning the weights in such ‘‘autoencoder’’ networks, but this works wellonly if the initial weights are close to a good solution.

我们描述了一种初始化权值的有效方法，这种方法允许深度自动编码器网络学习低维编码，这种编码作为一种降低数据维度的工具，其性能优于主成分分析。

We describe an effective way ofinitializing the weights that allows deep autoencoder networks to learnlow-dimensional codes that work much better than principal components analysisas a tool to reduce the dimensionality of data.

降低维度有助于高维数据的分类、可视化、通信和存储。

Dimensionality reduction facilitates theclassification, visualization, communication, and storage of high-dimensionaldata.

一种简单而广泛使用的方法是主成分分析（PCA），它在数据集中找到最大方差的方向，并用每个方向上的坐标来表示每个数据点。

A simple and widely used method isprincipal components analysis (PCA), which finds the directions of greatestvariance in the data set and represents each data point by its coordinatesalong each of these directions.

论文及完整源码下载地址：

http://page2.dfpan.com/fs/1lcaj2a210296160fe1/

更多精彩文章请关注微信号：