3 strokes to break the boundaries of Machine Learning Engineer
OpenCV school yesterday
This paper of about 1791 words, reading takes about 5 minutes;
Data collection system see the end of the text;
Keywords: artificial intelligence, machine learning, deep learning, mathematics, study recommendations
01. What are the boundaries of machine learning engineer is?
Most of the things are bounded. That machine learning boundaries, what is it?
Yes, that's mathematics. Mastery of the underlying mathematical machine learning, not only to deepen the understanding of the algorithm, but also more at ease in the model optimization stage.
02. How to break the boundaries?
I hope you can give the following reference content ideas.
• mathematical foundations of Machine Learning
The method is recommended: The practical significance and interest given the seemingly boring after learning as much as possible Sike minimum necessary knowledge.
The figure is an example using logistic regression to determine whether a male is the son of a suitable.
Wherein, Y = w1 * Height + w2 * character + w3 * wealth + w4 * color value + w5 * can be expressed as a polynomial Y = w1 * x1 + w2 * x2 + w3 * x3 + w4 * x4 + w5 * 5, after the adoption of Sigmoid function, the boys could be transformed for the probability of issues outstanding in-law.
Here it involves the sigmoid function, derivation algorithm, gradient descent, is the key to control over-fitting mathematical knowledge. Have problems to solve the problem, Sike critical point, it will not drill into a dead end, into a sea of mathematical knowledge alone.
• Machine Learning Linear Algebra
The method is recommended: Linear Algebra and try to combine practical significance
For linear algebra, it is the key to understanding and machine learning are: understanding of linear algebra and clever coupling of the real world.
Figure above the x, y dimensions can represent two things are observed.
x, y each of which has a large number of linear combination means that two dimensions of things a lot of different opinions,
These views will be calculated by machine, induction, deduction, and a combination of a large number of individual views, pulled out of line with most of the balance to arrive at universal conclusions. Is not that a wonderful event coupled clever thing -
• probability of statistical machine learning
The method is recommended: try the statistical probability seemingly tasteless topics of interest to combine knowledge and understanding
Hidden Markov (HMM) is a probabilistic graphical model algorithm in machine learning, study but also the difficulties of many algorithms job interview. Wang know almost from konjac, it clearly explains the principle of hidden Markov (HMM) algorithm with a scene of a game.
https://www.zhihu.com/question/20962240/answer/33614574 (link)
I was a soldier, practicing the three fighting shape, were furious state, the normal state and defense state. At the same time I also three passive skills, which are common flat A, crit (double attack damage), vampire (The Life).
Probability I shot crit in a rage state was 80%, play vampire probability is 5%;
at equilibrium shape, a ratio of a critical strike is 30%, the probability of hit vampire is 20%;
In defense form, crit the probability of success of 5%, vampire probability of 60%.
To sum up, the probability of Warrior in different states can play skills are not the same.
Originally, this career soldier in the rage state, there will be a circle around the red halo; when the state defense, there will be blue halo around. But now I'm playing a game, the game suddenly out of a bug: There are a silly x programmers to change the code of the game, he wrote to collapse, and soldiers from around the aura can not see. I could not see through the aura of the foot soldiers in the state know what burst.
Saying that now the question is: Since you do not see the foot of the aura, I can only estimate, "Warrior" in what state the explosion; but now I fight a boss, cut 10 times, eight are found crit, blood rushing to turn times out, you think my soldiers burst is most likely what state?
So, to understand the problem of probability by esoteric topics of interest, it is a lot easier.
