可视化自编码器训练结果

From Ufldl

Jump to: navigation, search
Line 30: Line 30:
原文中的“in particular”,应为强调之意。
原文中的“in particular”,应为强调之意。
原文“as some non-linear feature”中的“some”,似不译为好。非线性特征当然可以有很多,而这里计算出来的当然也只是其中一种,其意不言自明。
原文“as some non-linear feature”中的“some”,似不译为好。非线性特征当然可以有很多,而这里计算出来的当然也只是其中一种,其意不言自明。
-
 
+
:
:【原文】:
:【原文】:
By displaying the image formed by these pixel intensity values, we can begin to understand what feature hidden unit <math>\textstyle i</math> is looking for.
By displaying the image formed by these pixel intensity values, we can begin to understand what feature hidden unit <math>\textstyle i</math> is looking for.
Line 41: Line 41:
:【三校】:
:【三校】:
当我们用上式算出各像素的值、把它们组成一幅图像、并将图像呈现在我们面前之时,隐藏单元i所追寻特征的真正含义也渐渐明朗起来。
当我们用上式算出各像素的值、把它们组成一幅图像、并将图像呈现在我们面前之时,隐藏单元i所追寻特征的真正含义也渐渐明朗起来。
 +
:
:【原文】:
:【原文】:
If we have an autoencoder with 100 hidden units (say), then we our visualization will have 100 such images---one per hidden unit. By examining these 100 images, we can try to understand what the ensemble of hidden units is learning.
If we have an autoencoder with 100 hidden units (say), then we our visualization will have 100 such images---one per hidden unit. By examining these 100 images, we can try to understand what the ensemble of hidden units is learning.
Line 51: Line 52:
:【三校】:
:【三校】:
假如我们训练的自编码器有100个隐藏单元,可视化结果就会包含100幅这样的图像——每个隐藏单元都对应一幅图像。审视这100幅图像,我们可以试着体会这些隐藏单元学出来的整体效果是什么样的。
假如我们训练的自编码器有100个隐藏单元,可视化结果就会包含100幅这样的图像——每个隐藏单元都对应一幅图像。审视这100幅图像,我们可以试着体会这些隐藏单元学出来的整体效果是什么样的。
 +
:
:【原文】:
:【原文】:
When we do this for a sparse autoencoder (trained with 100 hidden units on 10x10 pixel inputs<sup>1</sup> we get the following result:
When we do this for a sparse autoencoder (trained with 100 hidden units on 10x10 pixel inputs<sup>1</sup> we get the following result:
Line 64: Line 66:
“一个稀疏自编码器”中的“一个”不必译出来。当然是一个,还能是几个?
“一个稀疏自编码器”中的“一个”不必译出来。当然是一个,还能是几个?
[[Image:ExampleSparseAutoencoderWeights.png|thumb|400px|center]]
[[Image:ExampleSparseAutoencoderWeights.png|thumb|400px|center]]
 +
:
:【原文】:
:【原文】:
Each square in the figure above shows the (norm bounded) input image<math>\textstyle x</math>that maximally actives one of 100 hidden units. We see that the different hidden units have learned to detect edges at different positions and orientations in the image.
Each square in the figure above shows the (norm bounded) input image<math>\textstyle x</math>that maximally actives one of 100 hidden units. We see that the different hidden units have learned to detect edges at different positions and orientations in the image.
Line 76: Line 79:
:【三校说明】:
:【三校说明】:
我想这里不宜用“对应”,因为图中小方块是10x10方阵排列的,但隐藏单元不是。怎么对应?逐行对应还是逐列?这都是未知的。因此为严谨起见,还是改为“某一个”较好。
我想这里不宜用“对应”,因为图中小方块是10x10方阵排列的,但隐藏单元不是。怎么对应?逐行对应还是逐列?这都是未知的。因此为严谨起见,还是改为“某一个”较好。
 +
:
:【原文】:
:【原文】:
These features are, not surprisingly, useful for such tasks as object recognition and other vision tasks. When applied to other input domains (such as audio), this algorithm also learns useful representations/features for those domains too.
These features are, not surprisingly, useful for such tasks as object recognition and other vision tasks. When applied to other input domains (such as audio), this algorithm also learns useful representations/features for those domains too.
Line 86: Line 90:
:【三校】:
:【三校】:
显而易见,这些特征对物体识别等计算机视觉任务是十分有用的。若将其用于其他输入域(如音频),该算法也可学到对这些输入域有用的表示或特征。
显而易见,这些特征对物体识别等计算机视觉任务是十分有用的。若将其用于其他输入域(如音频),该算法也可学到对这些输入域有用的表示或特征。
-
 
+
:
:【专业术语对照表】:
:【专业术语对照表】:
为了在后期校对时,使前后章节专业术语翻译统一,在此将本章中专业术语翻译的中英文对照总结到下表,以便统一修改,或用于后期专业名词附录。
为了在后期校对时,使前后章节专业术语翻译统一,在此将本章中专业术语翻译的中英文对照总结到下表,以便统一修改,或用于后期专业名词附录。

Revision as of 11:58, 7 March 2013

Personal tools