学术研究 —— 常用结论、说法

学术研究 —— 常用结论、说法

  • Multimedia:video, speech and music
  • non-linear ⇒ linear
  • 直接求解(较困难)⇒ iteration methods

    • 将会变得是分困难,此时可以使用子空间迭代的方式,


      其中的 记为 remainer;

  • 有正有负,进行统一处理(比如衡量误差)时,一般首先执行平方操作,以避免加和时的相互抵消;
    • 方差;
    • MSE;(均方误差)

1. 凸函数 & 非凸函数

  • Convex functions are attractive because they can be more reliably minimized than non-convex functions.
  • However, non-convex penalty functions can lead to enhanced sparsity of sparse signals.

2. dimensionality reduction

  • Ideally, the reduced representation should have a dimensionality that corresponds to the intrinsic dimensionality of the data

3. explanatory

  • Real-world data, such as :
    1. speech signals,
    2. digital photographs,
    3. fMRI scans

4. 数据的来源(data acquisition )

  • 传感器:RFID, robot & cameras