Machine Learning Free Labs

Supervised learning. If you are hearing or reading this term for the first time, then it may be completely unclear what it means. Don't worry. In this lab, you will get a comprehensive understanding of supervised learning; and, in the next chapter of the experiment, you will learn to use supervised learning to complete data prediction.

Linear regression is a relatively simple but important machine learning method. Mastering the principle of linearity and its solving methods is one of the fundamental requirements of handling linear regression. In addition, linear regression is also the cornerstone of the supervised learning regression. We hope that you can grasp some important ideas of machine learning through this experiment.

In the chapter of linear regression, we discussed the code implementation in detail by giving examples of housing prices in Boston. In this challenge, you need to use the relevant knowledge learned from linear regression to predict housing prices in Beijing.

Logistic Regression is a very basic classification method in machine learning. Thanks to simplicity and efficiency, the algorithm has been widely used in practical scenarios. In this course, we will explore the principle of logical regression and its algorithm implementation, and use scikit-learn to construct a logistic regression classification prediction model.

Among the classification predictions, there are few algorithms based on probability theory. Naive Bayes is one of them. The naive Bayesian algorithm is simple to implement and the efficiency of predictive classification is very high. Generally, naive Bayes is a very common algorithm. This chapter mainly explains the principle of naive Bayes algorithm from Bayes' theorem and parameter estimation, and combines it with data to implement it. Finally, we will present a practical application.

We have implemented linear regression using three methods. This challenge will require you to complete the polynomial regression fitting using TensorFlow and draw the fitting curve.

Linear regression is the simplest problem in machine learning. At the same time, linear regression is closely related to artificial neural networks. In this lab course, we will implement linear regression by using the API provided by TensorFlow 2.

We used linear regression as an example to demonstrate new changes and new features of TensorFlow 2. In fact, the more important role of deep learning frameworks such as TensorFlow is to build artificial neural networks. So, this experiment will use TensorFlow 2 to build a simple feedforward neural network.

An entire process of deep learning consists of four steps: model training, model evaluation, model saving and model deployment. In fact, all the contents we have mentioned before are focused on the first two steps. In this lab course, we will learn how to save and restore model in TensorFlow 2.

In the previous two weeks of the course, we have learned what supervised learning is and how different supervised learning methods are. The next week, you will fully understand unsupervised learning and learn to use unsupervised learning to perform data clustering.

We introduced the use of TensorFlow. You are supposed to be familiar with the concepts of tensor, session and computational graphs. In this challenge, we will try to build a multi-layer neural network structure using TensorFlow. Don't worry; there is no deep neural network involved here and the focus is on getting familiar with the use of TensorFlow.

We got familiar with the algorithms commonly used in regression and classification in supervised learning. Next we will lead you to learn unsupervised learning: clustering. In this experiment, we first explain the most commonly used centroid-based clustering method, and then introduce the most representative K-Means algorithm and other members of the family in detail.

We have learned how to build neural networks using TensorFlow, and demonstrated how to use low-level APIs and TensorFlow's Keras. This challenge course requires you to complete the construction of a shallow neural network independently.

We have already learned how to use TensorFlow to implement linear regression and build a shallow neural network. In this lab course, we will focus on deeper neural network construction, choosing VGG - the classic convolutional neural network structure as an example to achieve Fashion-MNIST classification task. In addition, the course will also teach us how to use pre-trained models for transfer learning.

From this chapter, we will start the journey of deep learning. The key to deep learning lies in the construction of deep neural networks. If you build your own deep neural network from zero, the process will be very complicated. Therefore, in order to implement the deep learning model more conveniently, we need to study the use of some common frameworks. Currently in the deep learning community, the more popular frameworks are TensorFlow, Keras, PyTorch etc., each with their own unique characteristics. Thanks to the support of Google and the activeness of the large developer community, TensorFlow is a reliable platform and updates quite frequently. Studying the use of TensorFlow will help you build deep learning models conveniently.