简体   繁体   English

跨时代的恒定准确性

[英]Constant accuracy over epochs

I am training a gan and I am the accuracy doesn't change over epoch meanwhile the loss is deacresing.我正在训练一个 gan 并且我的准确性不会随着时间的推移而改变,同时损失正在减少。 Is there something wrong or it is normal because it's a gan?是有什么问题还是正常的因为是gan?

Thank you in advance.先感谢您。

In order to fully answer this question (specific to your case) we'd need to know what loss function you are using and how you measure accuracy.为了完全回答这个问题(特定于您的情况),我们需要知道您使用的是什么损失函数以及您如何衡量准确性。

In general, this can certainly happen for a variety of reasons.一般来说,这肯定会由于各种原因而发生。 The easiest reason to illustrate is with a simple classifier.最简单的说明原因是使用简单的分类器。 Suppose you have a 2-class classification problem (for simplicity) and an input $x$ and label (1, 0), ie the label says it belongs to class 1 and not class 2. When you feed $x$ through your network you get an output: $y=(p_1, p_2)$.假设你有一个 2 类分类问题(为简单起见)和一个输入 $x$ 和标签 (1, 0),即标签说它属于第 1 类而不是第 2 类。当你通过你的网络输入 $x$你得到一个输出:$y=(p_1, p_2)$。 If $p_1 > p_2$ then the prediction is correct (ie you chose the right class).如果 $p_1 > p_2$ 那么预测是正确的(即你选择了正确的类别)。 The loss function can continue to go down until $p_1=1$ and $p_2=0$ (the target).损失函数可以继续下降,直到 $p_1=1$ 和 $p_2=0$(目标)。 So, you can have lots of correct predictions (high accuracy) but still have room to improve the output to better match the labels (room for improved loss).因此,您可以有很多正确的预测(高精度),但仍有改进输出以更好地匹配标签的空间(改进损失的空间)。

声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.

 
粤ICP备18138465号  © 2020-2024 STACKOOM.COM