Thank Didi very much for giving me so many interview opportunities. I also sincerely admire the quality of Didi interviewers. I hope everyone will stop being black. I also don't know why I have the courage to cast all the algorithms of large factories. However, the recent state has always been to do written tests, do evaluations, run interviews, cast all kinds of companies, and run interviews around like go
The Irises classification
1, Download and Install Class Libraries for Various Aspects of Machine Learning in Python
2. Import data and analyze through descriptive analysis, visualization and other data.
3. Create six models and select the model with the highest accuracy
The Irises dataset features:
1. All feature data are numbers
2. This is a classification problem, which can be easily solved by supervised learning algorithm.
There are two traditional classification methods: supervised classification and unsupervised classification.Supervised classification requires a priori knowledge of the classification of the area to be classified, i.e. the training area of all kinds of ground objects to be distinguished should be selected from the studied area for establishing the discrimination function.Commonly used supervised classification methods include: K nearest neighbor method, Mahalanobis distance classification, maximum likelihood method, etc.Supervision classification methods mainly include: mean value, method, etc.
Welcome to reprint, but please indicate the original source and author information. Introduction to Tensor Flow (1) Basic Usage refer to: http://wiki.jikexueyuan.com/project/tensorflow-zh/get_started/[email protected]: [email protected]: 2017-02-25 This example is mainly based on tensorflow's Chinese documents to learn the basic usage of tensorflow.According to the documentation, there are some main problems: 1. is the usage of Session () and InteractiveSession ().The latter uses Tensor.eval () and Operation.run () instead of Session.run (). Most
Keras application is an available depth learning model with pre-training weights.These models can be used for prediction, feature extraction and fine tuning. Weight is automatically downloaded when instantiating the model and stored in ~/.keras/models/ Available Models The image classification models with pre-training weights on ImageNet are: -Xception -VGG16 -VGG19 -ResNet50 -InceptionV3 Xception model is only available in TensorFlow, because it relies on the SeparableConvolution layer.Other models include TensorFlow and Theano.
Before starting the interview, after reading Dr. Li Hang's Statistical Learning Method, you can manually push some basic formulas of commonly used machine learning models.LC didn't do it, slightly lower than medium.The interview strategy is probably to invest in some small and medium-sized enterprises first and then find someone to push in a big factory like Toutiao.This is basically the idea of gradient descent, so I started with the resume of inside initialize randomly, a small company.
Introduction Recently, a text classification experiment was carried out. This is an opportunity to apply some of the machine learning methods previously learned, which is hereby recorded. Problem Description There is no fixed format for the original data. The main content is the conversation records of customer and agent. Each conversation record may have 3 tags, corresponding to the first class (5 classes), second class (80 classes) and third class
Remove punctuation marks
python for Punctuation Removal
text=u''' Job Description LeTV Recruiting android Automation Test Engineer Job Responsibilities: 1. Responsible for the design of android super tv TV equipment automation solutions, automated testing and tool development;2, according to the characteristics of the project research stability testing technology, complete the development and testing of related tools;Skills requirements: 1. It is preferred to have mobile terminal automation tool architecture design and developer, be familiar with Android automation tools Uiautomator, Monkey, MonkeyRunner 2, and have actual experience in writing automation test scripts in JAVA/Python language to test projects;3.
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