Report of Machine Learning
- An introduction to what is a rise up program, what is git, what is the logic in programming, and the programming basis for making software.For materials 'Get Started with Programming Fundamentals to Become a Software Developer', maybe I will explain a little. This material discusses the basics of programming and continues with designing the application you want to make.
Today I studied math for machine learning:
- Systems of Linear Equations
- Matrices
For matrices I just learned two things: - About matrices multiplication and
- What is the Identity Matrix.
Because discussing machine learning, of course the language used is Python. And here I learned some basic Python, Python is known as a programming language with the shortest syntax. Python is used because there are many libraries that are very helpful in machine learning
- Systems of Linear Equations
- For second weekly report,There is a lot materials and assignment.For materials,There is 3 core
- Using python to interact with the Operating System (literally interact with OS,in this case we use Linux)
- Crash course on python (learn about advanced python like substring,regex,etc)
- Introduction to Git and GIthub (How to manage your files locally and web server with github,we learn the basic command like pull,add,push,clone)
Today i learned : - Data Stream
Data stream can sending or receiving data without using callbacks.Data stream also process data piece by piece.So, using data stream can reduce our resources(memory) and it's better than load all data(it takes so much memory),the program will work easily or fast with minimum resource.The resources can used to another process.If you want build a program in real time acces or have a big data or have minimum resource, i suggest you to use data stream - Python Subprocess
Sub process in python mean that you can start new process or run new code.If you want to build an Automation,you can use Python Subprocess Module.Of course not only automation you can do,there is a lot things that you can do with sub process module.Because i am focusing the Machine Learning path,so i think i will use the Subprocess module to do Automation. - Error Handling
Same like other programming language,error handling in Python use 'Try and Except block',you write all codes in try and except in order to catch the error
- Using python to interact with the Operating System (literally interact with OS,in this case we use Linux)
Third week,i have larned about 'Data',there is 3 core in this week:
- Foundations: Data, Data, Everywhere (This we learn how all data interact with us in daily life)
- Ask Questions to Make Data-Driven Decisions (This,we start to learn how to use data we have collected)
- Prepare Data For Exploration (This,we learn about data type,manage data,filter,organized,and many other)
Today i have learned about :
- Data ethic and privacy
Talk about data,of course not all adata we can publish or process,there is a lot of data ethic and privacy.
I will take anonymization as example.We want our data is safe when company collect our data,so,what is anonymization?
Anonymization can eliminate/create initial.We often use anonymization at: Name,Telephone numbers,License plat and Etc
- Transformation data(Change structure data)
- Structured data (Have same types,usually created in rows and coloumns)
- Unstructured data (Different types)
- Wide data (By multiple coloumns) and Long data (By multiple rows)
- This week, we have learned 3 cores:
- Process Data from Dirty to Clean (How to do data cleaning, what thing we should clean dat)
- Analyze Data to Answer Questions (We learned queries, data formatting, data transform, etc)
- Share Data Through the Art of Visualization (We learn how to make data easy to understand to other people like stakeholders,client,etc)
Now I will explain a little bit more about 3 thing about
Process Data from Dirty to clean.
Why data must clean? We know even we have normalized a data, it is still not cleaned such duplicate ,missing data.So we must clean data to make sure the data valid.
Analyze Data to Answer Questions
In case we have a question, how many people using this product, we can use data from database like sql or excell, then enter the condition to get output
Share Data Through the Art of Visualization
Just like the word saying, we create visualization (graph, pie, etc) with a rule, in order to help other people what data is it, what is the biggest value, etc
- Process Data from Dirty to Clean (How to do data cleaning, what thing we should clean dat)
Period 11-15 September 2023 , i have learned about little bit about R programming language.
R programming language is like other programming language, but R programming more into statistic thing.
- R programming also have many advantages are:
- Open source
- Data manipulation, data visualization, and statistics packages
- Has many library you can use
Now, lets talk about Data structures at R programming language. Data structures at R programming language are List and Vector.
How we create a vector at R?
Use 'c(value1, value2, value 3)' . If the value is a number, you must add 'L' in each number, example c(1,9,10) To---> c(1L,9L,10L)
How we create a list at R?
Not much different from create a vector, just replace 'c()' function to 'list()' function.
- Open source
In this week, i have learned Mathematics for Machine Learning and Data Science Specialization. So, at this topic, I think i will explain a little bit about Linear Algebra
for Machine Learning and Data Science. Well, it's not much different with math at your previous class. There are 4 cores at this course but i will explain one.
- System of linear equation. As we know, there are two types System of linear equation. First one is System of linear equation 2 variables, and second one is System of
linear equation 3 variables. Of course all of them have matrices too, we know how to solve at math, but at this course teach us how we solve this problem at programming
In this week, i have learned about calculus, probability and statistics.
- At calculus we learn
- Derivatives and optimization (More advanced derivatives like euler number, logarithm natural)
- Gradient and gradients descent (gradient descent will used to train machine learning model with neural network)
- Optimization in neural networks and newton's method (talk about classification, and regression)
I will explain a little bit about the word 'Classification'
Classification divide two types. First one is Binary Classification (Only 2 output like yes / no, woman/man, and other). And another one is Multi classification ( More than 2 output like weather prediction sunny/rain/cloudy/storm and other)
- Derivatives and optimization (More advanced derivatives like euler number, logarithm natural)
- At probability and statistics we learn
- Probability theorem (naive bayes, bernouli)
- Sampling (Mean, variance, point estimation)
- Hypothesis testing
- In this week i have learn 3 cores of course are:
- Introduction to Machine learning (Talk about learning types in machine learning like supervised learning, unsupervised learning and reinforcement learning)
- Regression (Talk regression with multiple input variables, linear and logistic regression)
- Classification (how to solve use classification)
I have explained about first core, so i will explain a little bit about the other cores.
Regression
There are 2 regression. First one is linear regression (Used to estimate) and logistic regression (Just like classification). And there is multiple input variables as i know It evaluates the relative effect of variables on the dependent variable when holding all the other variables in the model constant.
Classification
There are 2 types, first is binary class (only 2 output like yes or no/man or women/ dog or cat, etc) and the other one is multi class (more than 2 output)
- Introduction to Machine learning (Talk about learning types in machine learning like supervised learning, unsupervised learning and reinforcement learning)
In this week, i have learned about Advanced Learning Algorithms. At Advanced Learning Algorithms, there are 2 cores .
Neural Network
Decision Tree
Now, i will explain a little bit about the core
Neural Network
Neural network have 3 important part, like another, input - process - output.
1. Input
2. Layers (This ,can 2 or more layers)
3. Output
Decision tree
Neural network have 3 important part
1. Root
2. Leaf root
3. Branches
So, what is difference between them?
Neural networks
1. All types data
2. Can solve hard problem
3. Hard and long time to train
Decision Tree
1. Focused on structured data
2. Only easy problem
3. Easy to train
In this week, i have learned about a more advanced in machine learning. The topics are :
Unsupervised Learning (The short meaning is how we can train unlabeled data)
Recommenders (How to build a system recommender)
Reinforcement Learning (machine learning training method based on rewarding desired behaviors and punishing undesired ones)
I will explain a little bit more (based on coursera)
Unsupervised learning
This course focus on clustering and anomaly detection. Clustering is process unlabeled data to create a group. Anomaly detection itself is the outlier in data, we can find anomaly with several algorithm like gaussian
Recommenders
Recommenders focus on filtering.
Reinforcement learning
This method is very often used to machine learning, the key is correct algorithm
For this week, i have learned
-Deep Learning
-Convolutional Neural Network
I will explain all point above
Deep learning
Deep learning is part of machine learning. Deep learning has a structure called ANN ( Artificial Neural Network) . ANN itself has 2 or more hidden layers to process the data and get the output. Computer vision also part of deep learning. I will explain a little bit about object detection. Object detection will scan object with 'square' shape, and detect inside. Most used real time detection is YOLO algorithm
Convolutional Neural Network
The most used in convolutional neural network are convolutional and pooling. Convolutional is get the 'abstract/core' in image, when pooling resize the image.
2 weeks before the end of the bangkit program, in this 19th week my team and I have completed a product capstone entitled EMPAQ. Since I have not yet described the capstone of my product, I will tell what EMPAQ is.
In short, EMPAQ is an application with the aim of mental health. EMPAQ helps users get services from peer counselors as well as professionals. But not all features can be accessed for free.
We only provide consultations with chatbots and peer counselors . To conduct a consultation with an expert, you will certainly have to pay. For the focus of our application is to minimize
suicidal ideation, if indicated suicidal ideation, it will be directed to the mobile service for further