QA Official

Using Bidirectional LSTM Model to Construct Chinese Word Segmentation

https://qaofficial.com/post/2019/05/03/24676-using-bidirectional-lstm-model-to-construct-chinese-word-segmentation.html 2019-05-03
Use Bidirectional LSTM for Chinese Word Segmentation Read Data This data is based on the 2014 People's Daily Marked Corpus and has been processed simply.Each behavior has a short sentence, and the word segmentation method is maximum segmentation, and the format is similar to: use | bidirectional LSTM| build | Chinese | word segmentation model | import numpy as np DATA_PATH = r'E:\python\demos\news-splite.txt' SENTENSE_NUM = 200000 #要读取

[Deep Learning Framework Keras] An Example of a Simple Convolution Neural Network

https://qaofficial.com/post/2019/05/03/24705-deep-learning-framework-keras-an-example-of-a-simple-convolution-neural-network.html 2019-05-03
​ 1. Constructing Model from keras import layers from keras import models model = models.Sequential() # input_shape指定输入的tensor为28*28*1,这里分别对应的是图片的三个维度的属性,即ima

[Deep Learning] News Topic Classification for Single-label Multi-classification Problems

https://qaofficial.com/post/2019/05/03/24695-deep-learning-news-topic-classification-for-single-label-multi-classification-problems.html 2019-05-03
# -*- coding: utf-8 -*- """单标签多分类问题之新闻主题分类.ipynb Automatically generated by Colaboratory. Original file is located at https://colab.research.google.com/drive/18TqrbGYm2J-jmR89KZHOa7vxbAr4eOz2 ### 问题解释 每个数据点只能划分到一个类别,但是类别总数大于2个,

[Experience] Getting Started with Deep Learning-Training and Testing Your Data Set

https://qaofficial.com/post/2019/05/03/24650-experience-getting-started-with-deep-learning-training-and-testing-your-data-set.html 2019-05-03
After several days of hard work, I successfully trained my own data set and tested a single picture. val accuracy is about 0.91 during training.It seems that the effect is still satisfactory, and whether it has been fitted has not been determined. In the training process, the most annoying thing is to handle the file path and file storage location. 1. ImageNet Classification Section: caffe is an example folder under the CAFFE (Convolutional Architecture for Fast Feature Embedding) model.

[Python Keras Combat] Quick Start: 30 seconds to get started with Keras

https://qaofficial.com/post/2019/05/03/24799-python-keras-combat-quick-start-30-seconds-to-get-started-with-keras.html 2019-05-03
1. Introduction to keras Keras is an advanced neural network API written in Python, which can run with TensorFlow, CNTK, or Theano as the back end.Keras's development focus is to support rapid experiments.It is the key to do a good job of research to be able to convert your ideas into experimental results with minimal delay. If you have the following requirements, please choose Keras: -Allows simple and fast prototype design (user-friendly, highly modular, extensible).

[Turn]' Zero Foundation Beginner-Level Deep Learning' Series Articles (Tutorial+Code)

https://qaofficial.com/post/2019/05/03/24788-turnamp#x27-zero-foundation-beginner-level-deep-learningamp%23x27-series-articles-tutorial-code.html 2019-05-03
Whether the coming big data era or artificial intelligence era, or the era when traditional industries use artificial intelligence to process big data on the cloud, as a programmer with ideals and pursuits, will he feel out immediately if he does not understand the ultra-hot technology of Deep Learning?Now it's time to save your life. The series of articles in "Zero Foundation Beginner Deep Learning" aims to help you, who loves programming, reach the entry level from zero foundation.

[data mining sklearn】knn solves three classification problems

https://qaofficial.com/post/2019/05/03/24625-data-mining-sklearnknn-solves-three-classification-problems.html 2019-05-03
Main Contents:1. Working Principle of knn2. knn development process3. Features of knn Algorithm4. Project Actual Combat: knn Realize Three Classifications of iris The Irises Data Set 1, KNN working principle1. Assume that there is a labeled sample data set (training sample set), which contains the corresponding relationship between each data and the classification to which it belongs.2. After inputting new data without labels, compare each feature of the new data

[iris] [keras] neural network classifier and [scikit-learn] logistic regression classifier construction

https://qaofficial.com/post/2019/05/03/24590-iris-keras-neural-network-classifier-and-scikit-learn-logistic-regression-classifier-construction.html 2019-05-03
original link: https://github.com/fastforward labs/keras-hello-world/blob/master/kerashelloworld.ipynb original title: "helloworld" inkeras All the code in this article is based on python2. The editor used is ipython notebook, which is an entry level. Advanced Neural Network Library enables developers to quickly build neural network models without worrying about the numerical details of floating-point operations, tensor algebra and GPU programming. Keras is an advanced neural network library, which is based on Theano or the backends

[machine learning] to write a fully connected neural network (3): classification

https://qaofficial.com/post/2019/05/03/24653-machine-learning-to-write-a-fully-connected-neural-network-3-classification.html 2019-05-03
Let's write a classification neural network without regularization in python.Traditional classification methods include clustering, LR logical regression, traditional SVM, LSSVM, etc.LR and svm are two classifiers, and multiple lrs or svm can be combined to form multiple classifiers.The multi-classification neural network uses softmax+crossthreshold to form the final multi-classification cost function j.Why to use this cost function may require knowledge of generalized linear models.Simply put, it is to maximize the entropy of the classification function.

classification of training data set imbalance problem processing

https://qaofficial.com/post/2019/05/03/24602-classification-of-training-data-set-imbalance-problem-processing.html 2019-05-03
什么是数据不均衡? 在分类中,训练数据不均衡是指不同类别下的样本数目相差巨大。举两个例子: ①在一个二分类问题中,训练集中class 1的样本数比