Geek time open class

If you're a student, or a junior development engineer just a few years out of school, or a data mining or analytics engineer, you might want to consider becoming a machine learning engineer. Because this position, not only high salary, and development prospects. If you're an engineer and your goal is to become an architect or a technical director, or a CTO for a company, then machine learning should also be a must for you. In the future, any Internet company with more than 500 million dollars in revenue will need machine learning capabilities. Machine learning is no longer an advanced technology. You should know that all Internet companies' business models are inseparable from advertising, recommendation and search, in which machine learning will play an important role. But is machine learning hard at all? How do I get started? What kind of machine learning engineers does the industry need? In this live broadcast, I will give you the definitive answer from my point of view. What are the requirements for math, English, and academic qualifications that you will obtain as a machine learning engineer? What is machine learning really doing at a macro level? What kind of machine learning engineer does the industry like? What is the best way to get started quickly and master machine learning?Copy the code

Introduction to machine learning algorithm engineers

Introduction to error

  • Bad state of mind
    • There is no free lunch
    • The barrier to entry is not high
    • High-paying industries inevitably bring a lot of competition
    • Technical people should focus on their own improvement
  • And master of none
    • Algorithm silly run a demo entry is easy to do things
    • The flood of junior talents and the lack of senior talents
    • Switching fields takes a lot of time. Don’t dabble in them. Don’t switch too often
    • The right way to identify areas to attack
  • Unclear goals
    • Running demo is not the same as getting started and writing a few lines of code is meaningless
    • Aim for people who can work
      • Complete projects independently
      • Don’t make stupid mistakes there are problems that can be seen in advance but not solved in advance and a week of running data is a waste of time
  • The lack of practice
    • A lot of data science is just empirical
    • The introductory material itself is prone to misinterpretation
    • The difficulties of getting things done may be more valuable than understanding the principles
      • Setting up the environment, small problems with memory bursting may seem simple but can be time consuming
      • Simple things, simple questions don’t panic
    • It is not practical to wait until you have a “complete understanding” of an area
      • Data is the accumulation of experience and the industry is constantly discovering new things that can’t wait to get in the car first
    • The correct method as soon as possible practical operation refinement, deep do not much
  • Confusion between competence and knowledge
    • Knowledge does not equal ability
    • Basic ability to solve the problem of ability (come up to do) can only test the accumulation of practice
    • Core competencies Programming data Data Science (SVM derivation)
      • There is a need for long-term, planned improvement
  • An introduction to difficult
  • Core competence
    • Valuable directions accelerate learning
  • Programming ability
    • Master at least one programming language, Python /R
    • Familiar with data processing methods and tools: pandas, PyTorch
    • Efficient implementation of machine learning algorithms
    • Performance tuning of models and algorithms
  • Mathematical ability
    • Master probability statistics and optimal theory
    • Mastering machines makes learning possible
  • Mathematical science related ability
    • System modeling methodology
    • Master the accuracy tuning methods of various models
    • Select the appropriate modeling scheme according to the actual data
    • Research cutting-edge technology to solve practical problems
  • The challenge of competency development
    • Ability does not equal knowledge
    • Do-it-yourself is better than training/course materials
    • There is a lack of advanced materials on the market
    • Do-it-yourself dilemmas: Where to start and what about problems

Fundamentals of Programming

  • Data processing and visualization
  • Python syntax sugar
  • profiler

Ii basic Mathematics

  • Basic machine learning concepts
  • Maximum likelihood estimation
  • Logistic regression and algorithmic model of hand lifting

Data science ability

  • model
    • Common model
    • Model tuning parameter
    • Model integration
  • data
    • Derived variables according to the business
    • Common Encoder variable selection strategies
    • How to build variables based on exploratory analysis and bad case analysis
  • Deep learning
    • Common network architecture implementations
    • The alchemy of deep learning