This week the task:
A, python-based preparations
1) installed Python development environment, PyCharm or so can Anaconda, according to personal habits preferences.
2) install the basic library, such as numpy, pandas, scipy, matplotlib
Second, this week, video learning content: https://www.bilibili.com/video/BV1Tb411H7uC?p=1
1) P4 Python basis
2 P1 Introduction to Machine Learning)
Concept: Machine learning is AI a branch, to design a computer system, according to data provided by a certain way of learning, with the increase in the number of training, we can continue to learn and improve in performance, learning model parameter optimization.
Category: Machine learning involves supervised learning, unsupervised learning and reinforcement learning ( 9'38 )
Supervised learning - through the existing data on the (x, y) to determine the new data (x) of the y value.
• Example: After several training children learn the concept of the moon, and then be able to judge whether something is a moon.
Unsupervised Learning - Analyzing the relationship between the data are not completely independent, P (X) P (Y) ≠ P (XY) . (Clustering)
• Example: thesaurus combination trained to get new words, new words get the probability of word combinations.
Action: 1 ) Cleaning Data / feature selection; 2 ) determining the arithmetic model / parameter optimization; 3 ) predicting the outcome ( 21 '00 )
[ ×] large data storage / parallel computing / Robot
[Difference] adopt certain rules when doing traditional methods ; the use of some of the rules is machine learning .
Multivariate linear regression model: constructing a plurality of predictive factors according to the value infinitely close to the actual value of the model.
Unlimited iterations so that the loss function (objective function) minimum, optimal model.
Machine Learning general process: data collection → data washing → Engineering feature (feature selection, parameter adjustment) → data → prediction model ( 37 '39 )
[Note] The actual amount of work to clean and features a large engineering
Machine learning methods:
1 ) using various algorithms to classify the data
Linear SVM / RBF SVM / Decision Tree / Naive Bayes / Linear Discrimination / QDA / AdaBoast / Random Forest等
2 ) different malleable profile training model
3 ) loss function to make appropriate adjustments to achieve optimal prediction data
4 ) the size of the design model to adapt to different devices
Taylor formula - Prediction E X values / inspection Gini image coefficient (gradient descent method)
Γ function
Convex function
Soft-max return
The classical probability - Birthday Paradox / packing problem → entropy (chaos extent reflect - decision trees, random forest)
3. Job requirements:
1) Paste Python environment and pip list screenshots, look at everyone's readiness. Please will not have the conditions for the development of the reasons and plans.
2) Paste video study notes, requires real, not plagiarism, handwriting can take pictures.
3) What is machine learning, what classification? With case, write your understanding.
Two-dimensional array directly using Unique deduplication will become the first two-dimensional array of one-dimensional array then de-emphasis, does not meet the demand. Therefore, a method of: further converted into an imaginary number deduplication.
Method 2: two-dimensional array and then converted into a set of tuples
Stacking np.stack () , according to axis different effects of different stack.
Matrix Multiplication
Multiplying the corresponding element