Machine learning Exercises (2) __ Machine learning

1. Foreword

Or the study is not solid ah, if there are any questions welcome to the message. Parsing may not be right, because I pushed it myself (covering my face). 2. Exercise 1 (convolution and Chihua)

Enter a picture size of 200x200, followed by a layer of convolution (kernel size 5x5,padding 1,stride 2), pooling (kernel size 3x3,padding 0,stride 1), and another layer of convolution (kernel Size 3x3,padding 1,stride 1), the output feature chart sizes are:

A. 95

B. 96

C. 97

D. 98

E. 99

F. 100

The correct answer: C (personal count is C, the reference answer later revised to C.) ）
Analytical:
First we should know the formula for calculating the size of the convolution or after the pool:
Outputw=⌊imagew+2padding−kernelsizestride⌋+1 Output_w=\lfloor \frac{image_w+2padding-kernel_{size}}{stride}\ Rfloor+1
Outputh=⌊imageh+2padding−kernelsizestride⌋+1 output_h=\lfloor\frac{image_h+2padding-kernel_{size}}{stride}\ Rfloor+1

Where padding refers to an outward-extending edge size, and stride is the length of the step, that is, each move.

This makes it much easier, first the long width is generally large, so we only need to calculate a dimension, so that after the first convolution size is:
200+2−52+1=99 \frac{200+2-5}{2}+1=99
After the first pool size is:
99+0−31+1=97 \frac{99+0-3}{1}+1=97
The size after the second convolution is:
97+2−31+1=97 \frac{97+2-3}{1}+1=97

The final result is 97. 3. Exercise 2 (SPSS basis)

In the basic analysis module of SPSS, the function is "to reveal the relationship between data in the form of row and column list" is ()

A. Data description

B. Related

C. Cross-table

D. Multiple corresponding

Correct answer: C 4. Exercise 3 (classification problem)

A prison face recognition access system is used to identify the person to be entered, and this system includes the identification of 4 different types of personnel: prison guards, thieves, food delivery, others. Which of the following learning methods is best for this application requirement: ().

A. Two classification issues

B. Multi-classification issues

C. Hierarchical Clustering issues

D. K-Center Point clustering problem

E. Regression issues

F. Structural analysis issues

Correct answer: B

Analytical:
Two categories: Each classifier can only divide the samples into two categories. The prison samples were warders, thieves, food-delivery officers, and others. Two classifications certainly won't work. Vapnik 95 proposed to the basis of the support vector machine is a two classification classifier, this classifier learning process is to solve a positive and negative two classification derived from an optimal programming problem (dual problem), to solve the multiple classification problem is to use the decision tree to classify two classifier cascade, The concept of VC dimension is to say the complexity of this matter.

Hierarchical clustering: Creating a hierarchical hierarchy to decompose a given dataset. Prison guards, thieves, food-delivery officers, or whatever, they should be equal in rank, so they can't. This method is divided into two ways: Top-down (decomposition) and bottom-up (merging).

K-Center Point clustering: Select the actual object to represent the cluster, each cluster using a representative object. It's a rule that divides around the center point, so it's not appropriate here.

Regression analysis: A statistical method for dealing with correlations between variables, where there is no direct relationship between prison guards, thieves, food handlers and others.

Structural Analysis: The structural analysis is based on the statistical grouping, calculates the proportion of each component, and then analyzes the internal structure characteristics of a general phenomenon, the overall nature, the overall internal structure in accordance with the time lapse of the statistical method of change regularity. The basic form of structural analysis is to calculate the structural index. It won't work here either.

Multi-classification problem: train several different weak classifiers for different attributes, then set them as a strong classifier. Here the prison guards, thieves, meal-giving and so on, respectively, according to their characteristics set basis, and then to distinguish between identification. 5. Exercise 4 (Logit and SVM)

About logit regression and SVM incorrect is ()

A.logit regression objective function is to minimize the posterior probability

B. Logit regression can be used to predict the size of the probability of event occurrence

C. SVM target is structural risk minimization

D.SVM can effectively avoid model cross fitting

Correct answer: A

Analytical:

A. Logit regression is essentially a method of maximum likelihood estimation of weights based on a sample, and the latter probability is proportional to the product of a priori probability and a likelihood function. Logit is only the maximum likelihood function, and does not maximize the posteriori probability, not to mention the minimization of the posteriori probability. And the minimization of the posterior probability is a simple Bayesian algorithm to do. A error

B. The output of the logit regression is the probability that the sample is a positive class, and the probabilities can be calculated and the correct

C. The goal of SVM is to find the hyperplane that makes the training data as separate as possible and the maximum classification interval should be the structural risk minimization.

D. SVM can control the complexity of the model through the regularization coefficient and avoid fitting. 6. Exercise 5 (SVM)

There are two sample points, the first point is a positive sample, its eigenvector is (0,-1), the second point is a negative sample, its eigenvector is (2,3), and the training set composed of the two sample points constructs a linear SVM classifier with the class equation of the classifier ()

A. 2x+y=4

B. x+2y=5

C. x+2y=3

D. 2x-y=0

Correct answer: C

Analytic: This problem simplifies, for two points, the maximum interval is Street, so find out Street can. 7. Summary

In this chapter, we mainly review the convolution and pooling of computational methods, SPSS base, classification problems, SVM and logit model. There are also a number of problems, you are welcome to correct the criticism.