快速看了一下，我感觉完全可以去掉独立的条件，但也许应该加上仅限于离散的情况。

#19 给机器学习预训练的模型发行一个 token ？让大家分享模型的同时可能还有些经济回报 hh ，最近正在想给 https://hub.autoai.dev/ 加这个功能。

给机器学习预训练的模型发行一个 token ？让大家分享模型的同时可能还有些经济回报 hh ，最近正在想给 https://hub.autoai.dev/加这个功能。

可以试试 EasyOCR

/go/promotions @Livid

@Livid go /t/promotions

go /promotion

> 有手写反向传播计算梯度的例子吗?因为看的只是推导,感觉还少一个像高数上那种一个定理后面跟个例子让我学习的.........还是一般情况下大家都是只了解下推理过程,最后计算机计算都是调方法?

其实就是 chain rule 反复用来用去。但我还真的曾经做了一个例子： https://tinynet.autoai.org/en/latest/induction/linear.html

这个例子只是 linear （或叫全连接？）层的，但是其他的基本上也通用。

> matlab,tensorflow,pytorh 等软件 /框架 /包是不是同一种神经网络(就是前面说的那种最简单的先正向计算结果,再反向计算梯度),实现原理是一样的?只不过其他的功能封装的有所不同?

据我所知，数学原理应该是一样的，但是优化的方法不同，例如 TF 1 会用静态图？我也不是很了解。

> 前馈神经网络是我前面说的输出=权重乘输入,没有计算梯度进行修改权重的方式,为啥还叫神经网络?个人理解应该要有纠正的环节,难道是权重手动瞎蒙吗?

感觉你的理解有误，前馈神经网络只是你的网络里没有环，但依然要反向传播算梯度。

> 如果理论没啥大问题,要开始实战,有没有大佬有什么推荐看的东西?

你可以自己实现一个神经网络框架。只有全连接的话很快。欢迎参考： https://github.com/autoai-org/Tinyml

Although the question is typically used metaphorically, evolutionary biology provides literal answers, made possible by the Darwinian principle that species evolve over time, and thus that chickens had ancestors that were not chickens,[4] similar to a view expressed by the Greek philosopher Anaximander when addressing the paradox.[3]

If the question refers to eggs in general, the egg came first. The first amniote egg—that is, a hard-shelled egg that could be laid on land, rather than remaining in water like the eggs of fish or amphibians—appeared around 312 million years ago.[6] In contrast, chickens are domesticated descendants of red junglefowl and probably arose little more than eight thousand years ago, at most.[7]

If the question refers to chicken eggs specifically, the answer is still the egg,[8] but the explanation is more complicated. The process by which the chicken arose through the interbreeding and domestication of multiple species of wild jungle fowl is poorly understood, and the point at which this evolving organism became a chicken is a somewhat arbitrary distinction. Whatever criteria one chooses, an animal nearly identical to the modern chicken (i.e., a proto-chicken) laid a fertilized egg that had DNA identical to the modern chicken (due to mutations in the mother's ovum, the father's sperm, or the fertilised zygote).[9][4][10][11].

It has been suggested that the actions of a protein found in modern chicken eggs may make the answer different.[10][11] In the uterus, chickens produce ovocleidin-17 (OC-17), which causes the formation of the thickened calcium carbonate shell around their eggs. Because OC-17 is expressed by the hen and not the egg, the bird in which the protein first arose, though having hatched from a non-reinforced egg, would then have laid the first egg having such a reinforced shell: the chicken would have preceded this first 'modern' chicken egg.[10][11] However, the presence of OC-17 or a homolog in other species, such as turkeys,[12] and finches[13] suggests that such eggshell-reinforcing proteins are common to all birds,[14] and thus long predate the first chickens.

@ztjryg4 谢谢支持！

@zthxxx 谢谢支持和 star! :)

@Latin 谢谢支持！

@ekidona 确实，现在还很少，而且有一些可能不能直接跑起来（因为格式调整过许多次），所以欢迎来提交模型 :) 将来也许会有私有仓库的形式。

@c0xt30a 感谢回复和 star，在我的设想里，这些引用的模型最好是作者自己上传的，由作者来控制 License，将来在把模型导入到我们的数据库的时候也会给作者选项让他来设置 License 。其他的我们能做的不多而且也挺复杂的 :(

另外看到老哥的 GitHub 上，似乎是在 EMPA ？我在隔壁的 ETH，有空也许可以来我们 office 交流交流哈哈 :)