自定义博客皮肤VIP专享

*博客头图:

格式为PNG、JPG,宽度*高度大于1920*100像素,不超过2MB,主视觉建议放在右侧,请参照线上博客头图

请上传大于1920*100像素的图片!

博客底图:

图片格式为PNG、JPG,不超过1MB,可上下左右平铺至整个背景

栏目图:

图片格式为PNG、JPG,图片宽度*高度为300*38像素,不超过0.5MB

主标题颜色:

RGB颜色,例如:#AFAFAF

Hover:

RGB颜色,例如:#AFAFAF

副标题颜色:

RGB颜色,例如:#AFAFAF

自定义博客皮肤

-+

用思想、知识和音乐去影响和改变

  • 博客(910)
  • 资源 (6)
  • 收藏
  • 关注

转载 keras 关于 dropout 的一点讨论

dropout in training and testing #5357 wenouyang commented on 11 Feb 2017In this link, devinplatt gives the following way to include dropout in training,model = Sequential()model.add(Dropout(0.5, inpu...

2018-03-26 15:08:22 6498

转载 model fine tune & get output by layer

从VGG19的任意中间层中抽取特征from keras.applications.vgg19 import VGG19from keras.preprocessing import imagefrom keras.applications.vgg19 import preprocess_inputfrom keras.models import Modelimport numpy as npbas...

2018-03-21 17:51:27 515

原创 Keras 常用callback function

TensorBoard通过结合TF的TensorBoard将网络结构以及运行时状态可视化from keras.callbacks import TensorBoardmodel.fit([encoder_input_data, decoder_input_data], decoder_target_data,          batch_size=batch_size,          epo...

2018-03-21 15:32:38 2078

转载 Keras同时用多张显卡训练网络

References.官方文档:multi_gpu_model以及Googlehttps://www.jianshu.com/p/db0ba022936f0. 误区目前Keras是支持了多个GPU同时训练网络,非常容易,但是靠以下这个代码是不行的。os.environ["CUDA_VISIBLE_DEVICES"] = "1,2"当你监视GPU的使用情况(nvidia-smi -l 1)的时候会...

2018-03-16 10:35:11 1448

转载 shell 终端字符颜色

[plain] view plain copy终端的字符颜色是用转义序列控制的,是文本模式下的系统显示功能,和具体的语言无关,shell,python,perl等均可以调用。   转义序列是以 ESC 开头,可以用 \033 完成相同的工作(ESC 的 ASCII 码用十进制表示就是 27, = 用八进制表示的 33)。     \033[显示方式;前景色;背景色m     显示方式:0(默认值)...

2018-03-15 20:20:14 641

转载 十分钟教程:用Keras实现seq2seq学习

原文:A ten-minute introduction to sequence-to-sequence learning in Keras 作者:Francois Chollet 译者:雁惊寒摘要:序列到序列学习(seq2seq)是一种把序列从一个域(例如英语中的句子)转换为另一个域中的序列(例如把相同的句子翻译成法语)的模型训练方法。目前有多种方法可以用来处理这个任务,可以使用RNN,也可以使...

2018-03-03 00:39:59 1649

转载 一文看懂 Attention 机制,你想知道的都在这里了。Enjoy

/* 版权声明:可以任意转载,转载时请标明文章原始出处和作者信息 .*/                                                     author: 张俊林       要是关注深度学习在自然语言处理方面的研究进展,我相信你一定听说过Attention Model(后文有时会简称AM模型)这个词。AM模型应该说是过去一年来NLP领域中的重要进展之一,在很...

2018-02-26 22:59:14 14658 7

转载 利用GAN生成MNIST-demo

转自:https://github.com/NELSONZHAO/zhihu/blob/master/mnist_gan/gan_mnist.ipynb本篇代码将使用GAN来学习MNIST数据集的生成,仅仅作为一个demo演示,后面会更新更加复杂的GAN模型In [2]:import tensorflow as tfimport numpy as npimport pickleimport ...

2018-02-23 16:46:05 4735

转载 循环神经网络(一般RNN)推导

本文章的例子来自于WILDMLvanillaRNN是相比于LSTMs和GRUs简单的循环神经网络,可以说是最简单的RNN。RNN结构RNN的一个特点是所有的隐层共享参数(U,V,W),整个网络只用这一套参数。RNN前向传导st=tanh(Uxt+Wst−1) ot=softmax(Vst)st为t时刻隐层的状态值,为向量。 ot为t时刻输出的值(这里是输入一个xt就有一个输出ot,这个是不必要的,...

2018-02-13 14:40:07 1003

转载 简单理解LSTM神经网络

Recurrent networks (Elman, 1990) are designed to model sequences, while recursive networks (Goller & Küchler, 1996) are generalizations of recurrent networks that can handle trees. ---Yoav Goldber...

2018-02-13 11:47:11 1321

转载 特征哈希(Feature Hashing)

转处:http://breezedeus.github.io/2014/11/20/breezedeus-feature-hashing.html#fn:fhash在特征处理(Feature Processing)中我介绍了利用笛卡尔乘积的方法来构造组合特征。这种方法虽然简单,但麻烦的是会使得特征数量爆炸式增长。比如一个可以取N个不同值的类别特征,与一个可以去M个不同值的类别特征做笛卡尔乘

2018-02-01 15:33:07 1180

转载 The Wide and Deep Learning Model(译文+Tensorlfow源码解析)

转处:http://blog.csdn.net/sxf1061926959/article/details/78440220?readlogAuthor: DivinerShi本文主要讲解Google的Wide and Deep Learning 模型。本文先从原始论文开始,先一步步分析论文,把论文看懂。再去分析官方开源的Tensorflow源码,解析各个特征的具体实现方法,以及模

2018-02-01 15:29:14 725

转载 TensorFlow Wide And Deep 模型详解与应用(二)

转处:http://geek.csdn.net/news/detail/235471作者简介:汪剑,现在在出门问问负责推荐与个性化。曾在微软雅虎工作,从事过搜索和推荐相关工作。 责编:何永灿([email protected]) 本文首发于CSDN,未经允许不得转载。TensorFlow Wide And Deep 模型详解与应用(一)前面讲了模型输入的特征,下面谈谈模型本

2018-02-01 15:27:17 1420

转载 TensorFlow Wide And Deep 模型详解与应用(一)

转处:http://geek.csdn.net/news/detail/235465作者简介:汪剑,现在在出门问问负责推荐与个性化。曾在微软雅虎工作,从事过搜索和推荐相关工作。 责编:何永灿([email protected]) 本文首发于CSDN,未经允许不得转载。Wide and deep 模型是 TensorFlow 在 2016 年 6 月左右发布的一类用于分类和回归的模型

2018-02-01 15:24:00 1657

转载 携程反爬中的Eleven参数-反爬与反反爬的奇技淫巧

http://www.shenjianshou.cn/blog/?p=302今天我们要聊点什么呢,之前说要聊去哪儿的,不过暂且咱们再放一放,先聊一聊去哪儿的干爹携程吧,上次我记得看了携程工程师霸气回应说懂爬虫的来去哪儿,懂反爬的来携程。我觉得特别棒,这种开放的心态和自信,正是一个开放的互联网环境所需要的。所以今天这节课虽然咱们以携程为例,但是我们还是以学习的目的

2017-12-02 16:32:30 5911 6

转载 【目标检测】RCNN算法详解

http://blog.csdn.net/shenxiaolu1984/article/details/51036677Girshick, Ross, et al. “Rich feature hierarchies for accurate object detection and semantic segmentation.” Proceedings of th

2017-10-25 19:19:58 370

转载 Selective Search for Object Recoginition

在前一段时间在看论文相关的工作,没有时间整理对这篇论文的理解。在前面的一篇博客【1】中有提到Selective Search【2】,其前期工作利用图像分割的方法得到一些原始区域(具体内容请查看【1】),然后使用一些合并策略将这些区域合并,得到一个层次化的区域结构,而这些结构就包含着可能需要的物体。      博客【3】已经有对这篇论文的一些简单介绍,写这篇博客不免有重复发明轮子之嫌,不想想

2017-10-24 10:49:03 314

转载 Faster-RCNN算法精读

论文:《Faster R-CNN: Towards Real-Time ObjectDetection with Region Proposal Networks》摘要:算法主要解决两个问题:1、提出区域建议网络RPN,快速生成候选区域;2、通过交替训练,使RPN和Fast-RCNN网络共享参数。 一、 RPN网络结构RPN网络的作用是输入一张图像,输出一批矩形候选区域,类

2017-10-23 17:12:34 361

转载 Faster RCNN详解:从region proposal到bounding box回归

网上很多关于Faster RCNN的介绍,不过这一片算是比较全的了,不仅包括整体流程、思想的介绍,也包括各个实现较为深入的介绍。大概内容记录如下(仅记录目前我感兴趣的部分):1 各种CNN模型以及数据库自从接触基于深度学习的目标检测这一领域以来,经常遇到各种CNN模型,比如ZF模型、VGG模型等等。同时也接触到各种数据集如PASCAL VOC、MNIST、ImageNet等等,博文

2017-10-23 17:07:35 6956 8

转载 目标检测方法简介:RPN(Region Proposal Network) and SSD(Single Shot MultiBox Detector)

最近几年深度学习在计算机视觉领域取得了巨大的成功,而在目标检测这一计算机视觉的经典问题上直到去年(2015)才有了完全使用深度学习的方法,也就是大名鼎鼎的Faster-RCNN以及和它同一时期的工作YOLO。目标检测就是要找到一张图中所有的物体和它们的位置,在这两篇文章之前,人们通常的做法是先用一些传统视觉的方法如selective search找到proposal,即比较可能是物体的一个

2017-10-23 16:21:39 556

原创 tensorflow简易安装

https://github.com/tensorflow/tensorflow下载whlpip install tf_nightly-1.head-cp27-none-linux_x86_64.whlpip install scipy -i http://pypi.douban.com/simple/pip install sklearn -i https://mirrors.u

2017-10-22 13:07:09 3464

原创 opencv python

void detectMultiScale(const Mat& image,CV_OUT vector& objects,double scaleFactor = 1.1,int minNeighbors = 3, int flags = 0,Size minSize = Size(),Size maxSize = Size());参数1:im

2017-10-19 20:40:46 325

原创 docker 安装 Cannot connect to the Docker daemon

安装: yum install docker-io   卸载:yum remove docker-io服务启动日志:/var/log/dockertime="2017-10-19T10:52:40.777968827+08:00" level=warning msg="You are running linux kernel version 2.6.32-431.23.

2017-10-19 11:02:10 611

转载 机器学习算法的调试 —— 梯度检验(Gradient Checking)

http://blog.csdn.net/lanchunhui/article/details/51279293http://ufldl.stanford.edu/wiki/index.php/%E6%A2%AF%E5%BA%A6%E6%A3%80%E9%AA%8C%E4%B8%8E%E9%AB%98%E7%BA%A7%E4%BC%98%E5%8C%96反向传播

2017-09-19 00:01:15 782

原创 java.lang.OutOfMemoryError: Direct buffer memory at java.nio.Bits.reserveMemory(Bits.java:658) at

java.lang.OutOfMemoryError: Direct buffer memoryat java.nio.Bits.reserveMemory(Bits.java:658)at java.nio.DirectByteBuffer.(DirectByteBuffer.java:123)at java.nio.ByteBuffer.allocateDirect(ByteBuf

2017-09-11 17:41:58 1938

转载 github ml code

38226vinta19267kerasModular neural network library based on Theano.14207SparkSpark is a fast and general engine for large-scale data processing.

2017-09-08 19:07:05 1803

转载 追踪 Netty 异常占用堆外内存的经验分享

https://zhuanlan.zhihu.com/p/21741364本文记述了定位 Netty 的一处漏洞的全过程。事情的起因是我们一个使用了 Netty 的服务,随着运行时间的增长,其占用的堆外内存会逐步攀升,大概持续运行三四天左右,堆外内存会被全部占满,然后只能重启来解决问题。好在服务是冗余配置的,并且可以自动进行 Load Balance,所以每次重启不会带来什么损失。

2017-09-04 19:57:01 8588

转载 Save and Load Your Keras Deep Learning Models

出自:https://machinelearningmastery.com/save-load-keras-deep-learning-models/Keras is a simple and powerful Python library for deep learning.Given that deep learning models can take hours,

2017-08-30 19:49:16 565

转载 在Docker下使用Nvidia GPU进行计算

http://blog.csdn.net/bychahaha/article/details/48493233要使用nvidia-docker来运行docker 容器nvidia-docker run -p 8503:8503 --mount type=bind,source=sentiment-phone_pre/tf_model,target=/models/sentiment-ana...

2017-08-29 20:50:21 1529

转载 executor.invokeAll和Future.get指定限制时间

设定时间的任务如果某一个子线程任务不能在规定的时间内完成,那么应该抛弃该任务,不再需要其结果,同时中断该任务,减少性能开销。Future.get方法有一个重载,传入限制的时间V get(long timeout, TimeUnit unit) throws InterruptedException, ExecutionException, TimeoutExceptio

2017-08-24 00:24:47 981

原创 netty 使用笔记

HttpObjectAggregator 这个handler就是将同一个http请求或响应的多个消息对象变成一个 fullHttpRequest完整的消息对象ChunkedWriteHandler 这个handler主要用于处理大数据流,比如一个1G大小的文件如果你直接传输肯定会撑暴jvm内存的,增加ChunkedWriteHandler 这个handler我们就不用考虑这个问题了,内部原

2017-08-15 00:17:13 272

转载 Convolutional Neural Networks卷积神经网络 Contents 一:前导 Back Propagation反向传播算法 网络结构 学习算法 二:Convolutional N

转自:http://www.gageet.com/2014/0878.phpContents一:前导 Back Propagation反向传播算法网络结构学习算法二:Convolutional Neural Networks卷积神经网络三:LeCun的LeNet-5四:CNNs的训练过程五:总结本文是我在20140822的周报,其中部

2017-07-31 16:38:43 527

原创 Data Augmentation & data Transfer Learning - TODO

TODO

2017-07-28 21:55:32 371

转载 Saving a Python dict to a file using pickle

Per Programming Python, 3rd Edition, there are a number of methods to store persistent data with Python:I often use flat files to read or write text (string) data using the os library.Flat files

2017-07-06 17:16:29 554

原创 [code review] lstm - sentiment analysis

Long short-term memory (LSTM) is a recurrent neural network (RNN) architecture (an artificial neural network) published in 1997 by Sepp Hochreiter and Jürgen Schmidhuber. Like most RNNs, an LSTM netwo

2017-07-06 17:12:30 965

转载 keras Embedding layer [Arguments input_dim: int > 0. Size of the vocabulary, i.e. maximum integer ]

keras.layers.embeddings.Embedding(input_dim, output_dim, embeddings_initializer='uniform',embeddings_regularizer=None, activity_regularizer=None, embeddings_constraint=None,mask_zero=Fal

2017-07-06 11:46:32 674

转载 lstm 做 文本的情感分析

github上可以参考的代码https://github.com/BUPTLdy/Sentiment-Analysis/blob/master/code/Sentiment_lstm.pyhttps://github.com/life-is-good/CommentFilter

2017-07-05 19:50:47 4328

转载 keras Usage of metrics 评价指标

Usage of metricsA metric is a function that is used to judge the performance of your model. Metric functions are to be supplied in the metrics parameter when a model is compiled.model.compile(

2017-07-05 19:50:32 10218

原创 docker keras 下 ImportError: cannot import name ctc_ops

Using TensorFlow backend.Traceback (most recent call last):  File "", line 1, in   File "/root/miniconda2/lib/python2.7/site-packages/keras/__init__.py", line 3, in     from . import activatio

2017-07-05 19:43:26 3819

转载 numpy文件存取-npz,npy

NumPy提供了多种文件操作函数方便我们存取数组内容。文件存取的格式:二进制和文本。二进制格式的文件又分为NumPy专用的格式化二进制类型和无格式类型。Numpy binary files (NPY, NPZ)load(file[, mmap_mode, allow_pickle, ...])Load arrays or pickled object

2017-06-28 10:56:58 103621

httpclient tutorial httpclient 指南

httpclient 指南 包括了详细的调用和常用代码 The Hyper-Text Transfer Protocol (HTTP) is perhaps the most significant protocol used on the Internet today. Web services, network-enabled appliances and the growth of network computing continue to expand the role of the HTTP protocol beyond user-driven web browsers, while increasing the number of applications that require HTTP support. Although the java.net package provides basic functionality for accessing resources via HTTP, it doesn't provide the full flexibility or functionality needed by many applications. HttpClient seeks to fill this void by providing an efficient, up-to-date, and feature-rich package implementing the client side of the most recent HTTP standards and recommendations. Designed for extension while providing robust support for the base HTTP protocol, HttpClient may be of interest to anyone building HTTP-aware client applications such as web browsers, web service clients, or systems that leverage or extend the HTTP protocol for distributed communication.

2018-03-08

mask rcnn paper

We present a conceptually simple, flexible, and general framework for object instance segmentation. Our approach efficiently detects objects in an image while simultaneously generating a high-quality segmentation mask for each instance. The method, called Mask R-CNN, extends Faster R-CNN by adding a branch for predicting an object mask in parallel with the existing branch for bounding box recognition. Mask R-CNN is simple to train and adds only a small overhead to Faster R-CNN, running at 5 fps. Moreover, Mask R-CNN is easy to generalize to other tasks, e.g., allowing us to estimate human poses in the same framework. We show top results in all three tracks of the COCO suite of challenges, including instance segmentation, bounding-box object detection, and person keypoint detection. Without tricks, Mask R-CNN outperforms all existing, single-model entries on every task, including the COCO 2016 challenge winners. We hope our simple and effective approach will serve as a solid baseline and help ease future research in instance-level recognition. Code will be made available.

2018-03-07

Applying Deep Learning To Answer Selection

Applying Deep Learning To Answer Selection- A Study And An Open Task

2018-03-07

Learning Phrase Representations using RNN Encoder–Decoder

Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation

2018-03-07

BPTT BackPropagation Through Time.pdf

BPTT paper This report provides detailed description and necessary derivations for the BackPropagation Through Time (BPTT) algorithm. BPTT is often used to learn recurrent neural networks (RNN). Contrary to feed-forward neural networks, the RNN is characterized by the ability of encoding longer past information, thus very suitable for sequential models. The BPTT extends the ordinary BP algorithm to suit the recurrent neural architecture.

2018-03-07

空空如也

TA创建的收藏夹 TA关注的收藏夹

TA关注的人

提示
确定要删除当前文章?
取消 删除