匠人_C-优快云博客

原创支持向量机(SVM)中对偶问题的理解

在硬间隔支持向量机中，问题的求解可以转化为凸二次规划问题： (1)(1)minw,b12||w||2(2)s.t.yi(wTxi+b)≥1,i=1,2,⋯,m." role="presentation" style="position: relative;">minw,b12||w||2s.t.yi(wTxi+b)≥1,i=1,2,⋯,m.(1)(2)(1)(1)(1

2017-12-26 15:19:52 3162

原创支持向量机

支持向量机（support vector machines，SVM)是一种二类分类模型。它的基本模型是定义在特征空间上的间隔最大的线性分类器，间隔最大使它有别于感知机；支持向量机还包括核技巧，这使它成为实质上的非线性分类器。支持向量机的学习策略就是间隔最大化，可形式化为一个求解凸二次规划的问题，也等价于正则化的合页损失函数的最小化问题。支持向量机的学习算法是求解凸二次规划的最优化算法。关于支持

2017-12-22 17:57:58 1027

原创决策树

决策树学习通常包括三个步骤：特征选择、决策树的生成和决策树的修剪。特征选择信息增益：“信息熵”（information entropy）是度量样本集合纯度最常用的一种指标。信息增益：假定离散属性a有V个可能的取值{a1, a2, …, av}, 若使用a来对样本集D进行划分，则信息增益计算公式为：我们可以使用信息增益来进行决策树的划分属性选择，ID3算法就是使用的该方法。增益率： a信息增

2017-12-21 11:07:35 2356

算法实例原地址：What is the expectation maximization algorithm? 现在有两个硬币A和B，要估计的参数是它们各自翻正面（head）的概率。观察的过程是先随机选A或者B，然后扔10次。以上步骤重复5次。如果知道每次选的是A还是B，那可以直接估计（见下图a）。如果不知道选的是A还是B（隐变量），只观测到5次循环共50次投币的结果，这时就没法直接估计A和B的正

2017-12-18 17:59:09 8425 2

原创机器学习好文收集

从最大似然估计开始，你需要打下的机器学习基石数学之美番外篇：平凡而又神奇的贝叶斯方法Bayes’ Theorem: The Holy Grail of Data Science理解矩阵（一）隐马尔科夫模型支持向量机系列...

2019-01-29 09:13:25 302

原创 jupyter notebook使用的python环境不对

对jupyter notebook进行一些操作后，在打开运行jupyter notebook时，发现我已经安装的一些python包无法导入了，如：import numpy as np报错：ImportError: No module named 'numpy'而我的numpy包是安装过的，比较奇怪。使用如下代码检测python环境：import sysprint(sys.exec...

2019-01-15 13:56:23 8795 1

原创 ImportError: cannot import name 'ensure_dir_exists'

jupyter notebook重装或更新后，执行jupyter notebook出现如下错误：Traceback (most recent call last): File &quot;/home/cc/install/anaconda3/bin/jupyter-notebook&quot;, line 7, in &amp;lt;module&amp;gt; from notebook.notebookapp im...

2018-12-29 17:02:57 2023

原创 YOLO系列（三）—— YOLOv3: An Incremental Improvement

本系列分为三篇，此文为本系列第三篇，其他文章：YOLO系列（一）—— You Only Look Once:Unified, Real-Time Object DetectionYOLO系列（二）—— YOLO9000:Better, Faster, Strongerproject: https://pjreddie.com/darknet/yolo/paper: YOLOv3: An I...

2018-12-12 11:39:24 944

原创 YOLO系列（二）—— YOLO9000:Better, Faster, Stronger

paper: YOLO9000:Better, Faster, Strongerproject: https://pjreddie.com/darknet/yolo/yolo v2在yolo的基础上做了一些改进，对yolo不熟悉的可以参考：YOLO系列（一）在paper中，并没有给出网络结构的完整示意图，下图是在网上找的，原网址查看更加清晰：https://ethereon.github.i...

2018-12-11 17:52:48 709

原创 YOLO系列（一）—— You Only Look Once:Unified, Real-Time Object Detection

paper: You Only Look Once:Unified, Real-Time Object Detectionproject: https://pjreddie.com/darknet/yolo/个人认为yolo最大的特点是在目标检测任务中实现了end-to-end的网络结构，不需要复杂的pipeline，实现了速度上的提升。大体的检测过程如下图所示：将图像分割成S×S的网格...

2018-12-11 09:57:55 798

原创 Multi-Person Pose Estimation

论文：Realtime Multi-Person 2D Pose Estimation using Part Affinity Fields（openpose）https://arxiv.org/pdf/1611.08050.pdfRMPE: Regional Multi-Person Pose Estimation（AlphaPose）http://mvig.sjtu.edu.cn/re...

2018-11-29 11:03:36 579

原创 Facial Expression Recognition

Deep Facial Expression Recognition: A Survey:https://arxiv.org/pdf/1804.08348.pdfImage based Static Facial Expression Recognition with Multiple Deep Network Learning:https://www.microsoft.com/en-us...

2018-11-27 15:26:06 3185

原创和为K的子数组

leetcode题目地址给定一个整数数组和一个整数 k，你需要找到该数组中和为 k 的连续的子数组的个数。示例 1 :输入:nums = [1,1,1], k = 2输出: 2 , [1,1] 与 [1,1] 为两种不同的情况。说明 :数组的长度为 [1, 20,000]。数组中元素的范围是 [-1000, 1000] ，且整数 k 的范围是 [-1e7, 1e7]。首先想到...

2018-11-06 10:17:16 1763

原创 NotFoundError: Op type not registered 'NonMaxSuppressionV3' in binary running on ...Make sure the...

使用tensorflow加载.pb文件时报如下错误：NotFoundError: Op type not registered ‘NonMaxSuppressionV3’ in binary running on cc-X299-AORUS-Gaming-3. Make sure the Op and Kernel are registered in the binary running in ...

2018-10-31 17:03:34 7509

原创 mnist = tf.keras.datasets.mnist mnist.load_data() ConnectionRefusedError

keras.datasets下载数据集时，由于文件是存储在亚马逊的服务器上，运行代码时一般会报如下错误：Downloading data from https://s3.amazonaws.com/img-datasets/mnist.npz--------------------------------------------------------------------------...

2018-08-29 10:20:56 15430 6

原创查看cuda版本

查看方法：nvcc --version若nvcc指令无法识别，先安装nvidia-cuda-tools：sudo apt install nvidia-cuda-toolkit在百度上搜索出来的查看方法是如下指令：cat /usr/local/cuda/version.txt个人感觉该方法有时是错误的，在我个人的机器上用上述指令得出的结果是CUDA Version ...

2018-07-19 09:33:06 14260

原创 ModuleNotFoundError: No module named 'graphviz'

import graphviz时出现如题的错误，且使用：conda install graphviz安装成功后依然出现上述错误。解决办法：pip install graphviz

2018-04-17 17:41:29 22706 14

原创 Kernel Restarting

Jupyter Notebook运行过程中, kernel died.　报错信息： The kernel appears to have died. It will restart automatically. 可能原因：安装的一些包现在可能已经不适用了，卸载并重新安装如下包： ipykernel ipython jupyter_client jupyter_core trait...

2018-04-09 10:41:35 12571

原创 ImportError: libcublas.so.9.0: cannot open shared object file: no such file or directory.

安装cuda 9.1,cudnn 9.1和tensorflow 1.6后，导入tensorflow时，报如下错误：ImportError: libcublas.so.9.0: cannot open shared object file: no such file or directory Failed to load the native TensorFlow runtime.错误原因...

2018-04-09 09:16:13 2208

原创 Tensorflow - ValueError: Parent directory of model.ckpt doesn't exist, can't save

使用tensorflow保存session实：saver.save(sess, "model.ckpt")出现如题所示错误，解决办法: 保存的路径改为相对路径saver.save(sess, "./model.ckpt")参考：Tensorflow - ValueError: Parent directory of trained_variables.ckpt doesn’t...

2018-04-04 10:12:15 5647 2

原创 include/darknet.h:25:43: fatal error: opencv2/highgui/highgui_c.h: No such file or directory

yolo9000工程使用opencv进行编译时出现如下错误： include/darknet.h:25:43: fatal error: opencv2/highgui/highgui_c.h: No such file or directory解决办法：需要安装opencv sudo apt-get install libopencv-dev参考：https://groups.g...

2018-03-21 10:52:19 7599 1

原创 SVM优化过程中，满足强对偶的理解

最近看了Free Mind写的支持向量机的系列文章：http://blog.pluskid.org/?page_id=683，感觉收获比较大。但是对于其拉格朗日对偶函数为什么满足强对偶条件并没有给出详细的解答。原文中是这样描述的：“如果原始问题是 Convex 的并且满足 Slater 条件的话，那么 strong duality 成立。需要注意的是，这里只是指出了 strong dualit

2018-01-22 16:32:12 1450

原创 python实现感知器算法

import numpy as npimport numpy.random as nrimport matplotlib.pyplot as pl%matplotlib inline# This notebook is based on an excellent tutorial by Kostis Gourgoulias (http://kgourgou.me/)# Specify

2018-01-11 14:15:25 4475

原创 UnicodeDecodeError: 'utf-8' codec can't decode byte 0xd0 in position 0: invalid continuation byte

使用pandas读取csv文件时：train = pd.read_csv('./data/d_train_20180102.csv')报如题所示的错误。解决办法：加编码格式如：train = pd.read_csv('./data/d_train_20180102.csv', encoding = "GB2312")

2018-01-08 16:28:27 8411

原创朴素贝叶斯

朴素贝叶斯法是基于贝叶斯定理与特征条件独立假设的分类方法。贝叶斯公式：P(A|B) = P(B|A) P(A)/ P(B)条件独立性假设：朴素贝叶斯使用特征条件独立假设的原因：个人理解是为了解决高维度带来的数据稀疏性。朴素贝叶斯有效的解释：有些独立假设在各个分类之间的分布都是均匀的所以对于似然的相对大小不产生影响；即便不是如此，也有很大的可能性各个独立假设所产生的消极影响或积极影响互相抵消，最终导致

2017-12-20 16:16:46 366

原创 K近邻算法

给定一个训练数据集，对新的输入实例，在训练数据集中找到跟它最近的k个实例，根据这k个实例的类判断它自己的类（一般采用多数表决的方法）距离度量：一般使用欧氏距离，也可以使用其他距离。 k值的选择：k较小，容易被噪声影响，发生过拟合。k较大，较远的训练实例也会对预测起作用，容易发生错误。分类决策规则：多数表决（应该也可以根据距离使用带权重的表决方式）。主要问题：如何快速的进行k近邻搜索。

2017-12-20 11:20:52 1483 2

原创感知机

感知机是二分类线性分类模型，其输入为实例的特征向量，输出为实例的类别，取+1和-1二值。感知机公式：感知机的几何解释是，线性方程w*x+b=0 将特征空间划分为正负两个部分：损失函数如果训练集是可分的，感知机的学习目的是求得一个能将训练集正实例点和负实例点完全分开的分离超平面。为了找到这样一个平面（或超平面），即确定感知机模型参数w和b，我们采用的是损失函数，同

2017-12-19 14:18:58 1017

原创 Anaconda 安装OpenCV

import cv2报错：ImportError: No module named 'cv2'解决办法：安装opencv。conda install -c menpo opencvpip安装方法应该类似。更多其他安装方法：How do I install Python OpenCV through Conda?

2017-12-15 10:12:52 32119 2

原创 tensorflow TypeError: run() got multiple values for argument 'feed_dict'

tensorflow在session调用run()时，报上述错误。import tensorflow as tfsess = tf.InteractiveSession()import numpy as npa = np.array([[1.0,2.0,3.0,4.0],[5.0,6.0,7.0,8.0],[9.0,10.0,11.0,12.0],[1.0,1.0,1.0,1.0]])w =

2017-11-15 17:58:10 6490

原创线性回归算法实现（最小二乘法，梯度下降）

一、最小二乘法import numpy as np;import matplotlib.pyplot as plt;初始化数据，网上随便找的数据，说是奥运会100用时，以及对应的年份。dataArray = np.array([[12, 1896], [11, 1900], [11, 1904], [10.8, 1908], [10.8, 1912],

2017-11-03 11:42:20 1866

原创检测tensorflow是否使用gpu进行计算

import tensorflow as tfsess = tf.Session(config=tf.ConfigProto(log_device_placement=True))查看日志信息若包含gpu信息，就是使用了gpu。其他方法：跑计算量大的代码，通过 nvidia-smi 命令查看gpu的内存使用量。

2017-10-29 23:10:56 98718 8

原创 Error parsing PCC subspaces from PCCT

安装ubuntu16.04时，出现错误 Error parsing PCC subspaces from PCCT ACPI Error: [SDS0] Namespace lookup failure, AE_NOT_FOUND (20160930?psargs-359) 个人解决办法：换ubuntu14.04进行安装。

2017-10-15 20:43:07 14194

原创 Ubuntu细节记录

1、右键打开终端设置：sudo apt-get install nautilus-open-terminal注销后生效。2、可执行.sh文件文件双击运行：直接在任意文件夹下，菜单中选择Edit->Preference->Behavior选项卡，可以看到默认为View executable text files when they are opened。把其选择为Ask each time

2017-10-15 20:36:11 446

原创 EOFError: Ran out of input

使用pickle.load(f)加载pickle文件时，报错：EOFError: Ran out of input. 可能原因：文件为空。解决办法：加载非空文件。其他解决办法： 1、加载前判断文件是否为空import osscores = {} # scores is an empty dict alreadyif os.path.getsize(target) > 0:

2017-10-10 16:15:20 95437 4

原创线性代数，矩阵相关学习资源推荐

教学视屏：线性代数 b站熟肉作品博客：理解矩阵（一）理解矩阵（二）理解矩阵（三）矩阵本质讲解 Jacobian矩阵和Hessian矩阵进阶知识

2017-09-20 14:59:25 1106

原创 Jupyter notebook中Python2和Python3共存

我之前Jupyter notebook中只有Python3，可以通过一下方式添加python2:sudo python2 -m pip install ipykernelsudo python2 -m ipykernel install --userpython2是你的系统上对应的python2的命令，我的python2的命令是python，故对应的命令是：sudo python -m pip i

2017-08-22 10:22:58 2553 2

原创 %matplotlib inline UnknownBackend: No event loop integration for 'inline'.

在ipython中执行%matplotlib inline报如下错误： UnknownBackend: No event loop integration for ‘inline’. Supported event loops are: qt, qt4, qt5, gtk, gtk2, gtk3, tk, wx, pyglet, glut, osx解决办法：使用jupyter notebook交互

2017-08-21 17:33:23 4540

原创 Jupyter Notebook 的快捷键

Jupyter Notebook 有两种键盘输入模式。编辑模式，允许你往单元中键入代码或文本，这时的单元框线是绿色的。命令模式，键盘输入运行程序命令，这时的单元框线是灰色。未选中的单元没有线框。两种模式都可用： Shift-Enter : 运行本单元，选中下个单元 Ctrl-Enter : 运行本单元 Alt-Enter : 运行本单元，在其下插入新单元命令模式 (按键 Esc 开启)：

2017-08-21 17:17:56 345

原创 Udacity Deep Learning 任务 1: notMNIST

预处理 notMNIST 数据，在此基础上训练一个简单的逻辑回归模型导入python库：# These are all the modules we'll be using later. Make sure you can import them# before proceeding further.from __future__ import print_functionimport m

2017-08-18 11:41:59 802

原创 You should consider upgrading via the 'pip install --upgrade pip' command.

使用pip3安装软件时报如下错误： You are using pip version 8.1.1, however version 9.0.1 is available. You should consider upgrading via the ‘pip install –upgrade pip’ command.解决办法：升级pip3：pip3 install –upgrade pip

2017-08-16 10:10:37 21081 1

ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design

Abstract. Currently, the neural network architecture design is mostly guided by the indirect metric of computation complexity, i.e., FLOPs. However, the direct metric, e.g., speed, also depends on the other factors such as memory access cost and platform characterics. Thus, this work proposes to evaluate the direct metric on the target platform, beyond only considering FLOPs. Based on a series of controlled experiments, this work derives several practical guidelines for efficient network de- sign. Accordingly, a new architecture is presented, called ShuffleNet V2. Comprehensive ablation experiments verify that our model is the state- of-the-art in terms of speed and accuracy tradeoff.

2018-12-03

An Extremely Efficient Convolutional Neural Network for Mobile Devices

Abstract We introduce an extremely computation-efficient CNN architecture named ShuffleNet, which is designed specially for mobile devices with very limited computing power (e.g., 10-150 MFLOPs). The new architecture utilizes two new operations, pointwise group convolution and channel shuf- fle, to greatly reduce computation cost while maintaining accuracy. Experiments on ImageNet classification and MS COCO object detection demonstrate the superior perfor- mance of ShuffleNet over other structures, e.g. lower top-1 error (absolute 7.8%) than recent MobileNet [12] on Ima- geNet classification task, under the computation budget of 40 MFLOPs. On an ARM-based mobile device, ShuffleNet achieves ∼13× actual speedup over AlexNet while main- taining comparable accuracy.

2018-12-03

Fine-Grained Head Pose Estimation Without Keypoints

Abstract Estimating the head pose of a person is a crucial prob- lem that has a large amount of applications such as aiding in gaze estimation, modeling attention, fitting 3D models to video and performing face alignment. Traditionally head pose is computed by estimating some keypoints from the tar- get face and solving the 2D to 3D correspondence problem with a mean human head model. We argue that this is a fragile method because it relies entirely on landmark detec- tion performance, the extraneous head model and an ad-hoc fitting step. We present an elegant and robust way to deter- mine pose by training a multi-loss convolutional neural net- work on 300W-LP, a large synthetically expanded dataset, to predict intrinsic Euler angles (yaw, pitch and roll) di- rectly from image intensities through joint binned pose clas- sification and regression. We present empirical tests on common in-the-wild pose benchmark datasets which show state-of-the-art results. Additionally we test our method on a dataset usually used for pose estimation using depth and start to close the gap with state-of-the-art depth pose meth- ods. We open-source our training and testing code as well as release our pre-trained models 1 .

2018-12-03

Focal Loss for Dense Object Detection

Abstract The highest accuracy object detectors to date are based on a two-stage approach popularized by R-CNN, where a classifier is applied to a sparse set of candidate object lo- cations. In contrast, one-stage detectors that are applied over a regular, dense sampling of possible object locations have the potential to be faster and simpler, but have trailed the accuracy of two-stage detectors thus far. In this paper, we investigate why this is the case. We discover that the ex- treme foreground-background class imbalance encountered during training of dense detectors is the central cause. We propose to address this class imbalance by reshaping the standard cross entropy loss such that it down-weights the loss assigned to well-classified examples. Our novel Focal Loss focuses training on a sparse set of hard examples and prevents the vast number of easy negatives from overwhelm- ing the detector during training. To evaluate the effective- ness of our loss, we design and train a simple dense detector we call RetinaNet. Our results show that when trained with the focal loss, RetinaNet is able to match the speed of pre- vious one-stage detectors while surpassing the accuracy of all existing state-of-the-art two-stage detectors.

2018-12-03

MobileNetV2: Inverted Residuals and Linear Bottlenecks

Abstract In this paper we describe a new mobile architecture, MobileNetV2, that improves the state of the art perfor- mance of mobile models on multiple tasks and bench- marks as well as across a spectrum of different model sizes. We also describe efficient ways of applying these mobile models to object detection in a novel framework we call SSDLite. Additionally, we demonstrate how to build mobile semantic segmentation models through a reduced form of DeepLabv3 which we call Mobile DeepLabv3. is based on an inverted residual structure where the shortcut connections are between the thin bottle- neck layers. The intermediate expansion layer uses lightweight depthwise convolutions to filter features as a source of non-linearity. Additionally, we find that it is important to remove non-linearities in the narrow layers in order to maintain representational power. We demon- strate that this improves performance and provide an in- tuition that led to this design. Finally, our approach allows decoupling of the in- put/output domains from the expressiveness of the trans- formation, which provides a convenient framework for further analysis. We measure our performance on ImageNet [1] classification, COCO object detection [2], VOC image segmentation [3]. We evaluate the trade-offs between accuracy, and number of operations measured by multiply-adds (MAdd), as well as actual latency, and the number of parameters.

2018-12-03

DSFD: Dual Shot Face Detector

Abstract Recently, Convolutional Neural Network (CNN) has achieved great success in face detection. However, it re- mains a challenging problem for the current face detection methods owing to high degree of variability in scale, pose, occlusion, expression, appearance and illumination. In this paper, we propose a novel face detection network named Dual Shot face Detector(DSFD), which inherits the archi- tecture of SSD and introduces a Feature Enhance Module (FEM) for transferring the original feature maps to extend the single shot detector to dual shot detector. Specially, Pro- gressive Anchor Loss (PAL) computed by using two set of anchors is adopted to effectively facilitate the features. Ad- ditionally, we propose an Improved Anchor Matching (IAM) method by integrating novel data augmentation techniques and anchor design strategy in our DSFD to provide better initialization for the regressor. Extensive experiments on popular benchmarks: WIDER FACE (easy: 0.966, medium: 0.957, hard: 0.904) and FDDB ( discontinuous: 0.991, continuous: 0.862) demonstrate the superiority of DSFD over the state-of-the-art face detectors (e.g., PyramidBox and SRN). Code will be made available upon publication.

2018-12-03

cascade r-cnn paper

In object detection, an intersection over union (IoU) threshold is required to define positives and negatives. An object detector, trained with low IoU threshold, e.g. 0.5, usually produces noisy detections. However, detection per- formance tends to degrade with increasing the IoU thresh- olds. Two main factors are responsible for this: 1) over- fitting during training, due to exponentially vanishing pos- itive samples, and 2) inference-time mismatch between the IoUs for which the detector is optimal and those of the in- put hypotheses. A multi-stage object detection architecture, the Cascade R-CNN, is proposed to address these prob- lems. It consists of a sequence of detectors trained with increasing IoU thresholds, to be sequentially more selec- tive against close false positives. The detectors are trained stage by stage, leveraging the observation that the out- put of a detector is a good distribution for training the next higher quality detector. The resampling of progres- sively improved hypotheses guarantees that all detectors have a positive set of examples of equivalent size, reduc- ing the overfitting problem. The same cascade procedure is applied at inference, enabling a closer match between the hypotheses and the detector quality of each stage. A simple implementation of the Cascade R-CNN is shown to surpass all single-model object detectors on the challeng- ing COCO dataset. Experiments also show that the Cas- cade R-CNN is widely applicable across detector architec- tures, achieving consistent gains independently of the base- line detector strength. The code will be made available at https://github.com/zhaoweicai/cascade-rcnn.

2018-12-03

深度学习最新中文版 pdf

深度学习书籍 2017年9月的最新高清pdf版， beta版第一章引言 1 1.1 本书面向的读者 . . . . . . . . . . . . . . . . . . . . . . . . . . 10 1.2 深度学习的历史趋势 . . . . . . . . . . . . . . . . . . . . . . . 11 1.2.1 神经网络的众多名称和命运变迁 . . . . . . . . . . . 12 1.2.2 与日俱增的数据量 . . . . . . . . . . . . . . . . . . . 17 1.2.3 与日俱增的模型规模 . . . . . . . . . . . . . . . . . . 19 1.2.4 与日俱增的精度、复杂度和对现实世界的冲击 . . . . 22

2018-01-15

How to Write makefile

makefile文件的编写，How to Write makefile.pdf 。全英文版本。

2015-07-07

空空如也

TA创建的收藏夹 TA关注的收藏夹

TA关注的人

ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design

An Extremely Efficient Convolutional Neural Network for Mobile Devices

Fine-Grained Head Pose Estimation Without Keypoints

Focal Loss for Dense Object Detection

MobileNetV2: Inverted Residuals and Linear Bottlenecks

DSFD: Dual Shot Face Detector

cascade r-cnn paper

深度学习 最新中文版 pdf

How to Write makefile

空空如也

深度学习最新中文版 pdf