Introduction-to-Machine-Learning
所属分类：人工智能/神经网络/深度学习
开发工具：TeX
文件大小：1745KB
下载次数：0
上传日期：2023-02-16 11:07:53
上 传 者：
sh-1993
说明： 机器学习导论，分享我在机器学习导论课程中遇到的理论和编程思想。笔记...
(Sharing both theoretical and programing ideas, that I came across at Introduction to Machine Learning course. notes, homework solution and python assignment)
文件列表:
.idea (0, 2023-02-16)
iml (407, 2023-02-16)
.idea\inspectionProfiles (0, 2023-02-16)
.idea\inspectionProfiles\Project_Default.xml (410, 2023-02-16)
.idea\inspectionProfiles\profiles_settings.xml (174, 2023-02-16)
.idea\misc.xml (226, 2023-02-16)
.idea\modules.xml (326, 2023-02-16)
.idea\other.xml (233, 2023-02-16)
LICENSE (1063, 2023-02-16)
Section1.0 (0, 2023-02-16)
Section1.0\KNN.py (3441, 2023-02-16)
Section1.0\plot 3.JPG (45852, 2023-02-16)
Section1.0\plot1.png (22182, 2023-02-16)
Section1.0\plot2.png (25893, 2023-02-16)
Section1.0\section_1.pdf (258517, 2023-02-16)
Section1.0\section_1.tex (6478, 2023-02-16)
Section2.0 (0, 2023-02-16)
Section2.0\Section2.pdf (289511, 2023-02-16)
Section2.0\Section2.tex (9073, 2023-02-16)
Section2.0\intervals.py (1839, 2023-02-16)
Section2.0\union_of_intervals.py (9514, 2023-02-16)
Section3.0 (0, 2023-02-16)
Section3.0\plot 3.1.png (30092, 2023-02-16)
Section3.0\plot 3.2.png (25415, 2023-02-16)
Section3.0\plot 3.3.png (15402, 2023-02-16)
Section3.0\section3.pdf (327371, 2023-02-16)
Section3.0\section3.tex (8454, 2023-02-16)
Section3.0\sgd.py (6245, 2023-02-16)
Section4.0 (0, 2023-02-16)
Section4.0\backprop_data.py (1983, 2023-02-16)
Section4.0\backprop_main.py (1752, 2023-02-16)
Section4.0\backprop_network.py (6456, 2023-02-16)
Section4.0\plot 4.1.png (63613, 2023-02-16)
Section4.0\plot 4.2.png (62488, 2023-02-16)
... ...
[](https://saarbk.github.io/iml)
# Introduction-to-Machine-Learning
Sharing notebook and theoretical ideas I came across at Introduction to Machine Learning course.
Any section including the relevant PDF soultions,  and py files
## 
(Warm-up)
\
[1.1](Section1.0/section_1.pdf) Linear Algebra
\
[1.2](Section1.0/section_1.pdf) Calculus and Probability
\
[1.3](Section1.0/section_1.pdf) Optimal Classifiers and Decision Rules
\
[1.4](Section1.0/section_1.pdf) Multivariate normal (or Gaussian) distribution
\
[Visualizing the Hoeffding bound.](Section1.0/plot1.png)
[k-NN algorithm.](Section1.0/KNN.py)
## [](https://github.com/saarbk/Introduction-to-Machine-Learning/blob/main/EX2/Section_2.pdf)
\
[2.1](Section2.0/Section2.pdf) PAC learnability of l2-balls around the origin
\
[2.2](Section2.0/Section2.pdf) PAC in Expectation
\
[2.3](Section2.0/Section2.pdf) Union Of Intervals
\
[2.4](Section2.0/Section2.pdf) Prediction by polynomials
\
[2.5](Section2.0/Section2.pdf) Structural Risk Minimization
[Union Of Intervals.](EX2/union_of_intervals.py)
Study the hypothesis class of a finite
union of disjoint intervals, and the properties of the ERM algorithm for this class.
To review, let the sample space be  and assume we study a binary classification problem,i.e. .
We will try to learn using an hypothesis class that consists of k disjoint intervals.
define the corresponding hypothesis as
=%5Cbegin%7Bcases%7D1%20&%5Ctext%7Bif%20%7D%20x%5Cin%20%5Bl_1,u_1%5D%5Ccup%20%5Cdots%20%5Ccup%20%5Bl_k,u_k%5D%20%5C%5C1%20&%5Ctext%7Botherwise%7D%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%20%5Cend%7Bcases%7D)
## 
\
[3.1](Section3.0/section3.pdf) Step-size Perceptron
\
[3.2](Section3.0/section3.pdf) Convex functions
\
[3.3](Section3.0/section3.pdf) GD with projection
\
[3.4](Section3.0/section3.pdf) Gradient Descent on Smooth Functions
[SGD for Hinge loss.](Section3.0/sgd.py)
In the file skeleton sgd.py there is an helper function. The function reads the examples labelled 0, 8
and returns them with the labels 1/+1. In case you are unable to
read the MNIST data with the provided script, you can download the file from [ Here](https://github.com/amplab/datasciencesp14/blob/master/lab7/mldata/mnist-original.mat).
_{hinge}=\max&space;(0,1-\mathbf{x}_i&space;y_i))
[SGD for log-loss.](Section3.0/sgd.py)
In this exercise we will optimize the log loss defined
as follows:
&space;=&space;\log(1+e^{-y\mathbf{w}\cdot&space;x}))
## 
\
[4.1](Section4.0/section_4.pdf) SVM with multiple classes
\
[4.2](Section4.0/section_4.pdf) Soft-SVM bound using hard-SVM
\
[4.3](Section4.0/section_4.pdf) Separability using polynomial kernel
\
[4.4](Section4.0/section_4.pdf) Expressivity of ReLU networks
\
[4.5](Section4.0/section_4.pdf) Implementing boolean functions using ReLU networks.
[SVM](Section4.0/svm.py)
Exploring different polynomial kernel degrees for
SVM. We will use an existing implementation of SVM, the SVC class from `sklearn.svm.`
[Neural Networks](Section4.0/svm.py)
we will implement the back-propagation
algorithm for training a neural network. We will work with the MNIST data set that consists
of 60000 28x28 gray scale images with values of 0 to 1.
Define the log-loss on a single example
%7D(W)=-%5Cmathbf%7By%7D%5Clog%5Cmathbf%7Bz%7D_L(%5Cmathbf%7Bx;%5Cmathcal%7BW%7D%7D))
And the loss we want to minimize is
=%5Cfrac%7B1%7D%7Bn%7D%5Csum_%7Bi=1%7D%5E%7Bn%7D%5Cell%20(%5Cmathbf%7Bx%7D_i,%5Cmathbf%7By%7D_i)(%5Cmathcal%7BW%7D)=%5Cfrac%7B1%7D%7Bn%7D%5Csum_%7Bi=1%7D%5E%7Bn%7D-%5Cmathbf%7By%7D_i%5Cast%20%5Clog%20%5Cmathbf%7Bz%7D_L(%5Cmathbf%7Bx%7D_i;%5Cmathcal%7BW%7D))
近期下载者：
相关文件：
收藏者：