TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Introducing Keras 2

445 点作者 mirceam大约 8 年前

13 条评论

minimaxir大约 8 年前
Copying my rare product endorsement from the previous submission:<p>Keras is so good that it is effectively cheating in machine learning, where even Tensorflow tutorials can be replaced with a single line of code. (which is important for iteration; Keras layers are effectively Lego blocks). A simple read of the Keras examples (<a href="https:&#x2F;&#x2F;github.com&#x2F;fchollet&#x2F;keras&#x2F;tree&#x2F;master&#x2F;examples" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;fchollet&#x2F;keras&#x2F;tree&#x2F;master&#x2F;examples</a>) and documentation (<a href="https:&#x2F;&#x2F;keras.io&#x2F;getting-started&#x2F;functional-api-guide&#x2F;" rel="nofollow">https:&#x2F;&#x2F;keras.io&#x2F;getting-started&#x2F;functional-api-guide&#x2F;</a>) will let you reverse-engineer most the revolutionary Deep Learning clickbait thought pieces.<p>It&#x27;s good to see that backward compatability is a priority in 2.0, since it sounds like a lot had changed.
评论 #13875091 未加载
评论 #13877938 未加载
评论 #13878052 未加载
juxtaposicion大约 8 年前
Will Keras2 support PyTorch as backend, in the future?<p>Answer: [0] No, there are no plans to support PyTorch. There is nothing to be gained in supporting every novelty framework that crops up every quarter. Our goal is to make deep learning accessible and useful to as many people as possible, and that goal is completely opposite to building up deep learning hipster cred.<p>[0]: <a href="https:&#x2F;&#x2F;github.com&#x2F;fchollet&#x2F;keras&#x2F;issues&#x2F;5299" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;fchollet&#x2F;keras&#x2F;issues&#x2F;5299</a>
评论 #13873098 未加载
评论 #13873696 未加载
caoxuwen大约 8 年前
Highly recommend this course - <a href="http:&#x2F;&#x2F;course.fast.ai&#x2F;" rel="nofollow">http:&#x2F;&#x2F;course.fast.ai&#x2F;</a> that utilizes Keras as the main programming tool
评论 #13873096 未加载
epberry大约 8 年前
Keras is fantastic. Not the tightest analogy and probably unoriginal but I think of it as the Python to Tensorflow&#x27;s C. It&#x27;s easy to drop into tensorflow flow when needed but you can probably get away with Keras for a long time. Also, Francois helped us when we DM&#x27;d him on Twitter which was incredible.<p>Thank you so much Francois! I&#x27;m incredibly excited about this release!
krick大约 8 年前
I&#x27;m only starting with all that machine-learning, NN stuff and as many others I want to ask for some guidance&#x2F;resources&#x2F;learning material. What I feel especially lacking is something very broad and generic, some overview of existing techniques (but not as naïve as Ng&#x27;s ML course, I assume). There exist a lot of estimators and classifiers, there exist a lot of techniques and tricks to train models, there exist a lot of details on how to design a NN architecture. So how, for instance, do I even decide, that Random Forest is not enough for this task and I want to build some specific kind of neural net? Or maybe I don&#x27;t actually need any of these fancy famous techniques, but rather there exist some very well defined statistical method to do what I want?<p>What should I read to start grokking this kind of things? I feel quite ready to go full &quot;DIY math PhD&quot; mode and consume some heavy reading if necessary, but where do I even start?
评论 #13873831 未加载
评论 #13874260 未加载
评论 #13873675 未加载
评论 #13875815 未加载
评论 #13873704 未加载
uptownfunk大约 8 年前
The mathematician in me has kept me from jumping into deep learning before I understand the mathematical and statistical underpinnings of the algorithms involved. Looking forward to reading through the latest book out by mit press and giving things a whirl with Keras which I&#x27;ve heard so much about.
评论 #13873252 未加载
评论 #13873276 未加载
gidim大约 8 年前
I love Keras but I think this update broke more things than you realized. For example it&#x27;s no longer possible to get the validation set score (val_acc) during training which renders early stopping impossible. This was a documented feature on your FAQ.<p>Is the old documentation still available? I&#x27;d like to wait before I upgrade.<p>Edit:typo
评论 #13873662 未加载
backpropaganda大约 8 年前
1. Still no support for multiple losses. Models like VAEs cannot be idiomatically implemented. The second loss has to be &#x27;hacked&#x27; in. Notice how in the official example for VAE, the kl_loss is computed using variables which are NOT available via the loss function (<a href="https:&#x2F;&#x2F;github.com&#x2F;fchollet&#x2F;keras&#x2F;blob&#x2F;master&#x2F;examples&#x2F;variational_autoencoder.py#L46" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;fchollet&#x2F;keras&#x2F;blob&#x2F;master&#x2F;examples&#x2F;varia...</a>)<p>2. It&#x27;s still an input-&gt;output paradigm, rather than a {input, output}-&gt;loss paradigm which gives more flexibility.<p>These two issues are the main reason why I stick to slightly lower level APIs, even though I _want_ to use Keras.
评论 #13873837 未加载
评论 #13873821 未加载
Kiro大约 8 年前
Is it better to learn Keras instead of tflearn?<p>Copying a comment I made in another thread where one response recommended Keras:<p>I currently have a small pet project where I think some simple ML would be cool but I don&#x27;t know where to start.<p>Basically my use case is that I have a bunch of 64x64 images (16 colors) which I manually label as &quot;good&quot;, &quot;neutral&quot; or &quot;bad&quot;. I want to input this dataset and train the network to categorize new 64x64 images of the same type.<p>The closest I&#x27;ve found is this: <a href="https:&#x2F;&#x2F;gist.github.com&#x2F;sono-bfio&#x2F;89a91da65a12175fb1169240cde3a87b" rel="nofollow">https:&#x2F;&#x2F;gist.github.com&#x2F;sono-bfio&#x2F;89a91da65a12175fb1169240cd...</a><p>But it&#x27;s still too hard to understand exactly how I can create my own dataset and how to set it up efficiently (the example is using 32x32 but I also want to factor in that it&#x27;s only 16 colors; will that give it some performance advantages?).
评论 #13874831 未加载
评论 #13874781 未加载
评论 #13881214 未加载
chestervonwinch大约 8 年前
Keras is a great wrapper library built upon two fantastic frameworks -- theano and tensorflow. I&#x27;m glad to see it is moving forward, and kudos to everyone involved in all these libraries!
diminish大约 8 年前
Slightly irrelevant but curious question about the Analytics for 7day (34K), 14day and 30day active users. I&#x27;m running a similar site so, could it be that, a lot of users reading documentation are using ad&#x2F;tracking blockers so that active users count appear higher than it actually is in GA. Documentation users tend to read quite high pages per session. If I&#x27;m right then they should see less page views per user than expected.
评论 #13876338 未加载
mooneater大约 8 年前
Awesome. Yet &quot;codebases written in Keras 2 next month should still run many years from now&quot; given that deep learning is no new, how can they that confident that this API will remain relevant years down the line?
评论 #13873284 未加载
rfeather大约 8 年前
I wonder why they decided to get rid of MaxoutDense. Is there something better or is it so trivial to implement they decided to drop it?
评论 #13873187 未加载