随笔分类 - Brain-NN
TensorFlow, CNN, Bayesian Deep Learning
摘要:不错的公开课,只是用的PyTorch。 Course Lessons Keras Prerequisites Change Keras Backend To Theano Preprocess Data For Training With Keras Create An Artificial Neu
阅读全文
摘要:Ref: https://github.com/tensorflow/models/blob/master/research/slim/nets/mobilenet_v1.md Keras Keras专门提供的API 一、模型API Keras doc: https://keras.io/appli
阅读全文
摘要:The Mask R-CNN was published March 2017, by the Facebook AI Research (FAIR). * Projects Using this Model. Mask R-CNN for object detection and instance
阅读全文
摘要:有这么几个类别 Ref: How to Choose Loss Functions When Training Deep Learning Neural Networks This tutorial is divided into three parts; they are: Regression
阅读全文
摘要:前言 一、学习资料下载 <21天实战caffe> <深度学习:21 天实战 Caffe> 随书代码资源 作者博客 二、官网 http://caffe.berkeleyvision.org/
阅读全文
摘要:读罢[UFLDL] ConvNet,为了知识体系的完整,看来需要实战几篇论文深入理解一些原理。 如下是未来博文系列的初步设想,为了hold住 GAN而必备的知识体系,也是必经之路。 [Paper] Before GAN: sparse coding [Paper] Before GAN: Zeile
阅读全文
摘要:主页介绍:https://research.googleblog.com/2017/09/build-your-own-machine-learning.html 结合:[TensorBoard] Cookbook - Tensorboard 安装指南: https://github.com/chr
阅读全文
摘要:Ref: Combining CNN and RNN for spoken language identification Ref: Convolutional Methods for Text [1] CONVOLUTIONAL, LONG SHORT-TERM MEMORY, FULLY CON
阅读全文
摘要:Ref: https://www.zhihu.com/question/66027838 Ref: http://www.cs.toronto.edu/~rgrosse/courses/csc321_2017/readings/L15%20Exploding%20and%20Vanishing%20
阅读全文
摘要:训练时的实时状态跟踪的重要性 不言而喻。 [Tensorboard] Cookbook - Tensorboard 讲解调节更新频率 直接上代码展示: 总之,不同的summary写入不同的writer对象中。
阅读全文
摘要:TF有两个scope, 一个是name_scope一个是variable_scope 第一个程序: 第二个程序: 输出为: 第三个程序: 可以看出:variable scope和name scope都会给op的name加上前缀 对比三个个程序可以看出: name_scope对 get_variabl
阅读全文
摘要:Ref: http://blog.csdn.net/u014595019/article/details/52759104 Time: 2min Successfully downloaded train-images-idx3-ubyte.gz 9912422 bytes. Extracting
阅读全文
摘要:From: Predicting Movie Review Sentiment with TensorFlow and TensorBoard Ref: http://www.cnblogs.com/libinggen/p/6939577.html Ref: https://machinelearn
阅读全文
摘要:Ref: http://blog.csdn.net/mebiuw/article/details/60780813 Ref: https://medium.com/@erikhallstrm/hello-world-rnn-83cd7105b767 [Nice] Ref: https://mediu
阅读全文
摘要:ResNet引入了残差网络结构(residual network),通过残差网络,可以把网络层弄的很深,据说现在达到了1000多层,最终的网络分类的效果也是非常好 Ref: http://blog.csdn.net/buyi_shizi/article/details/53336192 Ref: m
阅读全文
摘要:参考:https://www.zhihu.com/question/38102762 参考:CNN和RNN中如何引入BatchNorm 论文:Batch Normalization: Accelerating Deep Network Training by Reducing Internal Co
阅读全文
摘要:From: http://www.cnblogs.com/denny402/p/6932956.html [784, 10] fully connected w = tf.Variable(tf.truncated_normal([img_pixel_input, layersize], mean=
阅读全文
摘要:该系列主要是《Tensorflow 实战Google深度学习框架 》阅读笔记;有了Cookbook的热身后,以这本书作为基础形成个人知识体系。 Ref: [Tensorflow] Cookbook - The Tensorflow Way 第一章,简介(略) 第二章,安装(仅记录个别要点) Prot
阅读全文
摘要:CS231n Winter 2016: Lecture 5: Neural Networks Part 2 CS231n Winter 2016: Lecture 6: Neural Networks Part 3 by Andrej Karpathy 本章节主要讲解激活函数,参数初始化以及周边的知
阅读全文
摘要:From: https://github.com/jcjohnson/cnn-benchmarks#alexnet 先大概了解模型,再看如果加载pre-training weight。 关于retain这件事,插入231n的一页PPT。总之:数据多,筹码多,再大胆训练更多的weight;否则,别胡闹
阅读全文