5-Year Plan revise up (ver3.0)
Since Qiita’s article was on the CEO Tani, he was able to revitalize the five-year plan to become an AI engineer.
Thank you for the article.
Title: “Five-year plan ver3.0”
date: “2018-12-31 16:28”
Minimum knowledge required for AI engineers
First,I saw the minimum knowledge required by AI engineers in six parts.
Here, the future When doing business as an AI engineer, be sure to understand the overall picture of the basic knowledge of the six content.
(1) Programming skills
– numpy、pandas、matplotlib, Scikit. Translation-Learn TensorFlow and keras
Especially in this It is good to be proficient in pandas.
This is a convenient library for data preprocessing, although data preprocessing is essential for machine learning.
– Differential、Linear algebra, vectors, matrices, probabilities, etc.
(3) Knowledge of statistics
– Stdev, Variance, probability distribution, estimation, test, etc.
(4) Basic knowledge of machine learning
Supervised learning and unsupervised learning
pretreatment, feature design, learning and evaluation
Single regression, multiple regression analysis, least squares method, Perceptron, logistic regression
Decision tree, random forest, support vector machine, K-means
Implementation skills and knowledge of deep learning
Knowledge of frameworks such as TensorFlow and Keras.
The flow to make a model that has been studied in Scikit-learn.
1. Data collection and preprocessing of data completion of missing values and removal of outliers.
2. Feature design (selection of feature quantity)
3. Model development (model selection and learning)
4. Model evaluation… cross test, evaluation by mixing matrix, etc.
(5) Knowledge of working with databases using SQL。
-Select, insert, UPDATE, Delete, where, like, limit, SUM, AVG, Max, GROUP by, having, order by, table join, view, subquery, case, and so on.
(6) Cloud Knowledge
-Knowledge of cloud infrastructure such as AWS and GCP and Azure.
Large Although it may seem that there are many other than six, it is recommended to squeeze it into two (1) and (4) first of all at once.
The reason is that the program is actually written and visible, making it easier to learn.
Because it is self-taught and it is frustrated when entering from the theory first.
The recommended way to study artificial intelligence by yourself
I introduced the necessary knowledge in the previous section, but how do I learn them?
In order to learn from (1) to (6), we recommend that you acquire knowledge in the following sequence.
Phase1 python Learn about machine learning programming and artificial intelligence.
Phase2 machine learning programming.
Phase3 Challenge to Kaggle.
Phase4 SQL, scraping and cloud technologies.
Phase5 Use machine learning skills to produce products.
Phase6 Teach (maybe I don’t think this).
I believe that learning from each phase is the most rewarding and enjoyable way to learn.
Phase In 1, we refer to the programming beginners.
If you are new to programming, come to the phase Please read through the first.
Phase2, based on what we actually learned in Phase 1, we will explain how to study machine learning programming.
Already If you are doing machine learning programming using Scikit-learn, you may skip it.
Phase3 describes how to learn practical programming through competitions like Kaggle. (I don’t think so either)
Phase4 In doing machine learning, retrieving data from the database is often done. SQL Knowledge is mandatory. In addition to SQL, this section introduces techniques such as scraping (data collection) and cloud technology. (First from Access)
Phase5 Use machine learning skills to get your product production. Those who reach this level often learn through the product.
Phase6 Teaching six people may clarify what you didn’t know. So it is also a good idea to teach your friends about machine learning and deepen your understanding.
Since then, I will explain how to tackle these in each of the six phases, while introducing the recommended books.
# Development Environment Construction
・Ipython, Python, atom settings on your PC
・Atom debugging with Python alone
・Copy and paste the Web practice problem in WSL and Emacs
・Design and develop a Web top page
・Under the top page (index.html) developed demo.py demo.html Demo.css demo.js
・each demoxx.py design, demoxx.html design, Representative .css design
Standardized module names
Create a module list
2. Git, GitHub configuration-selection of collaborators
・Learning GitHub (Binder stuff)
3. Document Planning
・Create a module list
・Python Program Journal Organization
・HTML Journal Organization
・JS Program and Journal Organization
4. Development plan
1. Python acquisition plan
2. Actual Practice
3. Python Programming (Command prompt and on WSL)
5. Model Building
6. Ipyton Debug (Jupyternotebook on WSL) developed it under Atom Environment (NumPy, pandas Acquisition)
Ipython, Pyton, atom settings on PC
– Toshiba Dynabook Satellite B554/K use
-Test in development environment
– Blog up
# Git, GitHub configuration-selection of collaborators
-Request for collaborators
# Document Plan formulation
– Five-year Plan
-One year schedule
# reading Plan
-Qiita document conning (in progress)
-Why make object-oriented (purchased)
-New Linux textbooks (purchased-read after referring to the actual machine)
-Everyone’s python (purchased-in-progress, early)
-Nikkei Linux (purchase)
-Making from scratch deep Learning-theory and implementation of learning in Python (pre-purchased)
-Access Learning (create Address Book and learn about MySQL query)
-Copy and paste in Qiita document study (in progress)
-(Programming (Python), Statistical machine learning, deep learning (deep Learning), Mathematics for machine learning, probability and statistics)
-Easy to learn mathematics for understanding machine learning
-Ipython Interactive Computing and Visualization Cookbook (purchased-last)