Extensive theory, practice! This column is not an algorithmic derivation!

An overview of the

Classification of machine learning algorithm principle, in-depth analysis of decision tree, Bayesian algorithm, logistic regression, gradient descent, integrated learning, K-nearest neighbor, support vector machine, etc., and Python implementation source program, at the same time to share the practical oriented Kaggle advanced guide, the application of machine learning in Kaggle actual projects.

This column focuses on the application of machine learning algorithms in practical projects while explaining the principles of machine learning algorithms. It is divided into theoretical explanation and project practice (mainly Kaggle topics).

The article directories

  • An overview of the

  • Theory of article

    • Machine learning algorithm classification and program implementation
      • Convolutional Neural Network (CNN)
      • K -Nearest Neighbor, KNN
      • Support vector machine SVM
      • Hidden Markov Models (HMM)
      • Gradient Boosting learning
      • Logistic Regression
      • Activation function/loss function/normalization /SVM/BP/ random forest
    • A new approach to deep learning
      • NLP: The attentional mechanism
      • Figure neural network GNN

  • practice

    • Kaggle of actual combat
      • Kaggle: From the door to the ground
      • Project combat: method + source code
    • Actual combat experience

  • conclusion

Theory of article

Machine learning algorithm classification and program implementation

Logical regression, support vector machines, clustering, decision trees, Bayesian algorithms, etc. Each article will end with a general Python implementation source code for each algorithm. Considering the convenience for readers to study, take notes and review in time, each algorithm category is written as a long article. But you can bookmark it, and then an algorithm can be reviewed separately.

Convolutional Neural Network (CNN)

  • Machine Learning Algorithms — Approaching Convolutional Neural Network (CNN)
  • Machine learning algorithm — Convolutional Neural Network (CNN) principle explanation
  • Machine learning practical explanation (2) – | convolution neural network handwritten a convolutional neural network

K -Nearest Neighbor, KNN

  • Machine learning algorithm — K-nearest Neighbor (KNN) classification algorithm principle Explanation
  • Implementation of K-nearest Neighbor (KNN) classification algorithm in Python

Support vector machine SVM

  • Machine Learning Algorithms — Support Vector Machine (SVM)

Hidden Markov Models (HMM)

  • Machine learning algorithms — Hidden Markov Models (HMM) principle explanation and Python implementation

Gradient Boosting learning

  • Machine Learning Algorithm — Gradient Boosting

Logistic Regression

  • Machine Learning Algorithms — Logistic Regression

Activation function/loss function/normalization /SVM/BP/ random forest

  • In this paper, WE understand the Back Propagation method in neural network

  • Common activation function (excitation function) understanding and summary

  • Visualization of loss functions: A brief discussion on parameter space and regularization of models

  • Does CNN really need downsampling (upsampling)?

  • Explore further: Why feature normalization/standardization?

  • Top 10 amazing operations in convolutional neural networks

  • Don’t you know about optimization algorithms in machine learning? Now I’ve summed it up for you

  • Harbin Institute of Technology master students to achieve 11 kinds of data dimensionality reduction algorithm, code has been open source!

  • Machine learning interview 150 (2020) : SVM XgBoost feature Engineering

  • Facebook engineer to teach you what is a random forest, even if also can understand | zero base dry

Concept differentiation: The difference between machine learning, data science, artificial intelligence, deep learning and statistics!

In addition, neural network is a special branch of algorithms, so it will be listed separately and written in detail in the field of deep learning. The details are as follows:

  • Overview of different types of convolutional layers in neural networks

  • Machine Learning Algorithms — Approaching Convolutional Neural Network (CNN)

  • In this paper, WE understand the Back Propagation method in neural network

  • Machine learning practical explanation (2) – | convolution neural network handwritten a convolutional neural network

  • Does CNN really need downsampling (upsampling)?

  • Top 10 amazing operations in convolutional neural networks

  • An in-depth understanding of loss functions in computer vision

  • When support vector opportunity on neural network: This study reveals the relationship between SVM, GAN, Wasserstein distance…

  • Google has used its computing power to explode a paper that explains everything there is to know about the infinite width of the web

  • New network: graph neural network GNN

  • CVPR 2020: Huawei GhostNet, surpassing Google MobileNet, has been open source

A new approach to deep learning

After mastering basic machine learning algorithms, it is necessary to focus in the field of deep learning new methods, new branches, especially in NLP, a new method in the field of computer vision, some algorithms can play a good effect in the open field, this is also our project field, can consider the methods, such as: Transformer, geri weis-corbley, etc.

  • Summary of notes on 22 deep learning refining diagrams

NLP: The attentional mechanism

  • Deep learning | explanation Transformer (Attention Is All You Need.)

  • Deep learning | Batch Normalization principle and actual combat

  • Deep learning | is reading the Normalization model of deep learning

  • Text depth representation model — Word2vec & Doc2vec word vector model

  • Extract dialogue characters from a Dream of Red Mansions by Google BERT Chinese application

  • Out of Distribution(OOD) detection based on depth model is introduced

  • 7 cca shut & Radios | ACL 2020 award-winning paper; A review of Bayesian deep learning

Figure neural network GNN

  • The “natural graph network” message delivery method is proposed by Wellings team

  • Overfitting can also be recycled: some people used it to reconstruct 3D surfaces in high definition, reducing the parameters by 99%

  • Farewell to RNN, welcome to TCN! It’s time for the stock-market forecasting task to embrace new technology

practice

Kaggle of actual combat

Kaggle: From the door to the ground

  • Exclusive tips: Free Kaggle contest handouts!

  • Undergraduate promotion GM Record: Kaggle competition advanced skills sharing

  • Kaggle primer, this one is enough

  • What do I need a gold medal for: Can Kaggle’s results really give Google and FB a run for their money?

  • 23 k +, making the star from scratch the depth study of practical tutorial | PyTorch official recommendations

Project combat: method + source code

  • Machine learning practical explanation (2) – | convolution neural network handwritten a convolutional neural network

  • The application of machine learning field | logic return “Kaggle Titanic disaster”

  • Keras realizes CNN: Handwritten digit recognition accuracy 99.6%

  • Top 5 machine learning projects on GitHub

  • CV of actual combat | using OpenCV panoramic image stitching

  • CV of actual combat | use OpenCV implement road vehicle counting

  • Focus on fast machine learning training algorithms, UC Berkeley, You Yang 189 pages of doctoral thesis published

  • How to win two KDD Cup 2020 crowns in one season? Meituan advertising team open solutions

  • Kaggle X-ray pneumonia test match second program parsing | CVPR 2020 Workshop

  • Chinese forces occupy KDD: Won all the champions and runners-up of “Big Data World Cup”, Beihang won the best Student Paper award…

Actual combat experience

  • Why is your model so poor, and what are the techniques of deep learning tuning?

  • Machine learning paper reappears, here are five issues you need to pay attention to

  • Where do you find large data sets for machine learning?

Column recommended

-“Take you start computer vision project, step-by-step description of column | project recommendation, open source

– Computer vision | top paper study

conclusion

In the continuous update, if you encounter any problems in practice, welcome to communicate. If you reply “add group” in the public account, you can enter the Learning exchange group of Micro. ✉ ️ [email protected]

More projects are coming –>

Pay attention to the public account Mai Micro AI Institute, and learn with more fellow travelers. And don’t forget to subscribe!