QA Official

[Competition Sharing] Kaggle-Toxicomment [Keras Multiple Binary Classification, High Quality Comment Corpus, Use of Pre-trained Word Vector] 2019-05-06
summary recently watching a Kaggle competition [toxcomment] The goal of the competition is to judge whether a written comment is a poison comment or not. At the same time, poison comments are specifically divided into six categories.【’toxic’, ‘severe_toxic’, ‘obscene’, ‘threat’, ‘insult’, ‘identity_hate’】 This blog mainly shares the new postures learned. Keras can actually do multiple 2 classifications at the same time Baseline[0.051] implemented by Bi-LSTM actually does 6 2 classifications at the same time, but I didn't know that I could do this before!

macro-average and micro-average of Multi-label classification Performance Evaluation 2019-05-06
Generally, we use accuracy when evaluating the performance of the classifier. Consider in the context of multi-class classification accuracy = (Number of Samples Correctly Classified)/(Number of Samples Classified) In fact, it looks pretty good, but there may be a serious problem: for example, an opaque bag contains 1,000 mobile phones, including 600 iphone6, 300 galaxy s6, 50 Huawei mate7,50 and 50 mx4 (of course, these information classifiers are unknown.

summary of single-layer neural network, multi-layer perceptron and in-depth learning 2019-05-06
Some Thoughts on Neural Networks: Single Layer-Multilayer-Depth  Neural Network is essentially an approximator. An important basic attribute is the universal approximation attribute. Universal Approximation Attributes: In 1989, George Cybenko published an article "Approval by Super Positions of A sigmoidal Function". The article proves that under the condition of only a single hidden layer, any continuous and non-linear Sigmoidal function can fit any continuous function well as long as there are enough hidden layers.

23 Design Patterns-Agency Pattern "Intermediary Company" 2019-05-05
For things that have not been used for a long time, it is easy to forget.Baidu has not found what it wants for a long time.From today on, some things will be recorded, not only for your own convenience, but also to help others One cannot avoid dealing with intermediary, such as renting a house. You seldom meet the landlord, but the so-called middleman landlord, and so on...What's more, if

40+APIs and Libraries for Face Detection and Face Recognition 2019-05-05
facerecognition-stephen from ramda laboratory.Sample Code and Graphic Demo Click on Our API provides face recognition, face detection, eye positioning, nose positioning, mouth positioning, and gender classification.If you have any questions, just send an email to [email protected] (Detection)-Computer Vision Face Recognition and Face Detection.This is a perfect substitute for present, we have a free API for face detection.animetrics face recognition-animetrics' face recognition API can be used for face detection in pictures.

Centos7 Installs Python3.6 2019-05-05
Install dependencies first yum -y groupinstall "Development tools" yum -y install zlib-devel bzip2-devel openssl-devel ncurses-devel sqlite-devel readline-devel tk-devel gdbm-devel db4-devel libpcap-devel xz-devel Download python3 installation package wget Create python3 directory and unzip: mkdir /usr/local/python3 tar -xvJf Python-3.6.2.tar.xz compile installation cd Python-3.6.2 ./configure --prefix=/usr/local/python3 make && make install Create Connection Test Installation ln -s /usr/local/python3/bin/python3 /usr/bin/python3 # 在/usr/bin 目录下存

J2EE software design principles, commonly used design patterns-factory pattern, Singleton pattern, template method, Adapter pattern, composite pattern, strategy pattern, agent pattern 2019-05-05
Six Principles of Software Design (Code Enriching ing) 1. Open-close principle: open to expansion and close to modification 2, Liskov Substitution Principle: Subclasses must appear wherever accumulation occurs 3, single responsibility principle: Single function and responsibility, can only embrace one change 4, Dependence Inversion Principle: Relying on Abstraction, Not on Implementation 5, Interface Segregation Principle: Provide users with small interfaces, using multiple specialized interfaces is better than using a single multi-interface 6, Dimitri Principle: Try to have fewer relationships with non-friends # design patterns are divided into three categories:

Java Design Pattern _ Behavior _ Mediator Pattern _ Rent a House 2019-05-05
Please indicate the source for reprinting: article comes from: [Liang Jingming's blog] 1. Why is there an intermediary When you come to work in a strange city, it is inevitable to solve all kinds of life problems. For those of you who are unfamiliar with life, how to solve the problems is very important.If we can't solve the problem directly, please tell the intermediary to solve the problem through

LibFaceDetection Open Source Library Introduction and Use 2019-05-05
unsigned char * result _ buffer//buffer, which must be 0x20000 bytes in size.buffer memory for storing face detection results, !!its size must be 0x20000 Bytes!!Unsignedchar * gray _ image _ data//single channel gray image (y in YUV data)Intwidth//width of single-channel gray imageIntheight//height of single-channel gray imageIntstep//the step parameter of the single-channel gray image is the same as the width of the single-channel gray image, inputimage, itmustbegray (single-channel) image!Float scale//The scale

SIFT Feature Point Detection 2019-05-05
SIFT feature point detection algorithm is an algorithm for detecting local features. It obtains features and matches image feature points by finding feature points in a graph and their descriptors related to scale and orientation. SIFT algorithm has the following characteristics: 1.SIFT feature is a local feature of an image, which is invariant to rotation, scale scaling and brightness changes, and also maintains a certain degree of stability to view angle changes, affine transformation and noise.