Talking about How to Cultivate Students' Deep Learning in Primary School Mathematics Teaching
1. Use innovative thinking to promote deep learning. Innovative thinking is an important part of students' thinking consciousness. While students receive education, it is also a process of cultivating and exercising personal thinking. Under the teaching goals of the new era, it is more inclined to explore and develop students' own abilities and creative abilities. .
Primary school mathematics, with its typical and unique thinking training, has become one of the important courses in the stage of enlightenment education. The formation of mathematical thinking ability has a great impact on students' future cognition of the world and the establishment of mathematical thinking.
Mathematics is one of the most important subjects in elementary school. Compared with Chinese and English, the logical thinking requirements of mathematics are obviously higher than those of them. Strengthening the cultivation of mathematical logical thinking and innovative thinking is the key to truly promoting students' mastery of mathematical knowledge and improving their personal abilities. Therefore, it is particularly important to learn to innovate in learning.
Therefore, in the classroom that promotes their in-depth learning, it is necessary to cultivate innovative thinking ability, make students interested in thinking more, encourage them to think more, and purposefully and consciously intersperse problems related to life.
Let them think according to their own thinking, although sometimes they may find some unexpected answers and mistakes, which can be ignored, and finally answer this question with an active and correct answer, and give guidance to the way of thinking, thereby improving the overall quality of the students. innovative thinking ability.
Here we take the continuous addition operation within 10 as an example. For 1+3+4=8, students can not only calculate with (1+3)+4=4+4=8, but also calculate with 1+(3+4) =1+7=8 to calculate.
At the same time, it can also be calculated in various ways such as 1+3+4=(1+3)+4=4x2=(1+3)+(1+3)=(1+3)x2=8.
There may be many students who propose various calculation methods, and of course some are "clumsy". We must not criticize his thinking, but encourage him, saying that his thinking is absolutely correct, and let him change it. Take a look at other people's way of thinking and see if he likes it or not. Is that faster and easier? It doesn't matter if he insists on his own way of thinking. The speed is relatively slow and cumbersome, but we can hold game competitions in the classroom, answer in groups, and the fastest and most accurate can get a small prize or honor encouragement.
In this way, students will naturally change their own thinking, and get the comparison results of the two ways of thinking in their hearts, which will stimulate their enthusiasm for thinking and thinking about new problem-solving ideas, thereby promoting students' in-depth learning.
Another example is the teaching of the perimeter of a rectangle. You can start with the simplest steps, add the lengths of the four sides, and then find that the perimeter relationship of 2 times the long side plus 2 times the short side is left to the students themselves. The teacher can give appropriate guidance. It is also possible to allow students to actually measure by themselves, and then discover the rules from it, forming a training process of self-discovery.
In addition, in the calculation process of the perimeter, length and width can be multiplied by 2, or length multiplied by 2 plus width multiplied by 2 can be adopted. Try to let students feel the best plan by themselves, combined with the cultivation of innovative thinking, to promote students' in-depth learning of perimeter. 2. Increase practical operations and promote deep learning.
Hands-on operation is one of the important learning methods advocated by the new curriculum. Effective hands-on operation is the key to effective teaching, which can stimulate students' interest in learning, improve students' practical ability, and promote students' in-depth learning.
The theoretical knowledge students learn from books cannot reflect their real level and ability, and they need to deepen their understanding of theoretical knowledge through physical objects, so as to gain the fun of learning.
Practice has shown that students can more directly experience the joy of learning mathematics during the process of physical manipulation or experimental testing, promote students' in-depth learning, and lay a solid foundation for later learning and exploration.
For example: the lesson "Division with Remainder" in the second grade is a new lesson after learning complete division. How to let the children understand the meaning of division with a remainder? After many attempts before class, I chose to let the children do it by hand, so that they can experience why there is a division with a remainder.
How is the remainder obtained? And let the children experience the difficulty of this lesson by themselves, the remainder must be smaller than the divisor. The learning utensils I choose are medium-sized lentils. These beans are of moderate size and shape, and they are not easy to roll off. Operation 1: Put 8 beans on 2 plates on average.
The kids got the results quickly: 4 beans per plate. Operation 2: 9 beans are placed on 2 plates on average. Not long after starting to work, a child raised his hand. When communicating after the operation, he said: Teacher, this topic is wrong. I hurriedly asked him: What's wrong?
He said: After I finished dividing, there was one more bean. I asked: What should I do with the extra 1 capsule? Can we continue to divide? He said: can no longer be divided. The other kids also agreed with him. One more grain is called the remainder in mathematics, which can be expressed by the formula 9÷2=4 (grain)...1 (grain).
From the children's eyes, I can see their suddenness, as if they are saying: The title is correct, so it is like this.
By letting them operate the process of dividing beans by themselves, they can experience the origin of the remainder and discover problems by themselves. This is what deep learning refers to: "Under the guidance of teachers, students focus on certain challenging learning topics, actively Meaningful learning processes to engage, experience success, and develop".
Another example is the lesson "Area" in the third grade. Students do not have the concept of "area" in life, but students have the experience of comparing sizes. The whole class started with a dispute between Xiaopang and Xiao Dingding. They both thought that the graphics in their hands were big. Let the students be the little teachers to help Xiao Ding Ding and Xiao Pang solve this problem.
When the direct comparison cannot be operated, I provide some tools to the students, so that the children can use the tools to operate and draw conclusions. During the process, they actively participated and worked hard to come up with a good method.
In the final communication process, some children used small squares of paper of the same size to spread on two figures, and came to the conclusion that the figure with more small squares of paper would be bigger. Some children chose the transparent graph paper I provided to compare, whichever figure occupies more squares, whichever figure will be bigger.
These are all successful methods. During this process, the children experienced how to use their brains to find solutions, operate with their own hands, and finally successfully solve problems. This process is deep learning. It also paves the way for the subsequent knowledge of how to calculate the area of graphics.
Therefore, in the appropriate courses, increasing students' practical operations will continuously promote students' in-depth learning during the entire process of operation. 3. Use diversified teaching to promote deep learning.
For primary school mathematics, problem solving is an important part of primary school mathematics, and using multiple solutions to a problem is conducive to the construction of students' thinking and can also promote students' deep learning. It is very important for the development of students' thinking.
With the continuous implementation of the new curriculum reform in our country, the new curriculum standard clearly points out that a variety of methods should be used in solving problems, and in the teaching and practice of the subject of mathematics, the diversified teaching mode of solving problems is also vigorously promoted.
When we taught the horizontal and vertical calculations of "Abdication and Subtraction", I showed the scene picture of the bird tree and introduced it with the familiar scene of the children. There were originally 9 birds on the tree, but 5 birds flew away, how many birds are left now?
The children have already learned the subtraction of single digits, and the children can give the formula immediately: 9-5=4 (only). Then change the situation map: there were 12 birds on the tree, and 5 birds flew away. How many birds are left now? Who will do the math? Blackboard writing: 12-5=? Group discussion.
During the exchange, the children actively raised their hands and got various answers. Health 1: If there are 10 birds, 5 will fly away and 5 will remain. Now it is 12, which is 2 more birds than 10, so the answer is 7.
According to the student's idea, write on the blackboard: 10-5=5, 5+2=7. Student 2: If there were originally 12 birds, 2 flew away first, and 10 remained; 3 flew away, and 7 remained. I immediately asked: Do you mean to divide the small into two flights?
Can you tell me how it was divided again? Health 2: That is, divide 5 into 2 and 3. According to the meaning of Sheng 2, I wrote down the second small calculation method on the blackboard: 12-2=10, 10-3=7.
After student 2 finished communicating, student 3 raised his hand and said: You can also fly away one by one, counting backwards from 12, 11, 10, 9, 8, 7. The children's minds are colliding, and they are actively looking for methods based on the questions I raised, and they are constantly thinking in different methods.
This has achieved another main purpose of the "deep learning" ability training we want, which is to cultivate students' independent higher-order thinking ability. Fourth, do what you like and promote deep learning. interest is the best teacher. Choose to carry out in-depth learning according to the interests and hobbies of the students.
This is the strategy we need most. This can improve the motivation and efficiency of learning. Students are willing to learn, work, and devote their energy and time. The primary school teaching work is aimed at children aged 6 to 11.
Game teaching is very popular among students, and it is also widely used in elementary school mathematics teaching. It will make more abstract mathematical concepts interesting to the boring classroom. Primary school students are naturally active and lively, and it is their nature to love to play.
So in my teaching process, I still try my best to organize various small games and integrate the knowledge points of mathematics into it. I also think that making games is a very good way, so I often use it in the teaching process of teaching the first and second grades.
For example, the teaching focus in the third grade is to multiply and divide two or three digits with one digit. Once, I saw an exploratory article on the product of one digit multiplied by 99 and 999 in the Journal of Mathematics.
It reminds me of the simple calculation method that students like most like clever calculations. I happened to come across a question 6×99 in the exercise book. Students like to use vertical calculations. Then I asked them: Who can directly calculate the answer without vertical formula?
I have a method as ingenious as a clever calculation, and tomorrow we will compare whose method is the most ingenious. In the class on the second day, the children came up with many good methods, and several of them thought of the fast calculation method mentioned in the article, 6×99=6×100-6.
The kids agreed this was their favorite ingenious method. So I immediately asked what if you encounter 6×999? Can you quickly calculate the answer? As soon as I finished speaking, the little hands underneath were already raised high.
In fact, we should make the most practical exploration based on what we have learned, combined with realistic conditions, so as to ensure the feasibility, effectiveness, and operability of students' thinking about problems.
Guide students to conduct in-depth learning according to their interests, hobbies, and what they like, and stimulate students' enthusiasm for learning, exploring, and applying knowledge. I think this is also another theme in "deep learning": the disguised cultivation of practical problem-solving abilities.
They gain knowledge and exercise, as well as validation from themselves and others. Deep learning is based on the needs of students and pays attention to the perception of students' hearts. The learning process is not only visible, but also "touches the soul". Even after leaving the classroom and school, learners can still maintain a strong desire and ability to learn.
For the development of students, we should help students construct a learning experience that students need to make the indispensable classroom learning more valuable.
Google AI Writing Project: Neural Network Pseudo-Original
Talking about how to promote students' deep learning in primary school mathematics
How to carry out in-depth learning in the classroom is the key to the new round of curriculum reform. As mathematics, how to carry out in-depth learning is even more urgent. Combined with my teaching, I think it may be better to do this? 1. Pre-class preview is the implementation of in-depth learning The basic premise is good copywriting .
Let students learn before class, through reading and drawing knowledge points, clarify the basic content of the text knowledge, and understand the basic spirit of the text, which is the basis for improving students' acceptance of new knowledge and strengthening the achievement of key points of knowledge.
Then students who have enough energy to learn will carry out problem-solving exercises to consolidate, strengthen, and improve, strengthen their understanding and recognition of basic knowledge, and produce the same direction of reinforcement of the knowledge they have learned. This link is the key, to ensure the learning of basic knowledge, to ensure the proficiency of basic skills, and even strengthen them.
These works have laid the foundation for us to carry out deep learning, so that we can select points of interest and carry out deep learning. 2. Select topics for in-depth study according to the interests and hobbies of the students.
This is the strategy we need most. This can improve the motivation and efficiency of learning. Students are willing to learn, work, and devote their energy and time.
For example, when I was teaching students about the stability of triangles, I asked students to make their own triangles and quadrilaterals. In the case of the same material, try the problem that the graphic has stronger stability. Students make graphic objects of different materials , through the external force given to different objects, observe which figure is easy to deform?
Some students also took part in the comparison of circular objects. Finally, in the classroom communication, students arranged different objects of the same material, and the triangle was the most stable conclusion.
In fact, we should make the most practical exploration based on what we have learned, combined with realistic conditions, so as to ensure the feasibility, effectiveness, and operability of students' thinking about problems.
Guiding students to carry out in-depth learning and exploration according to their interests, hobbies, and realistic conditions can stimulate students' enthusiasm for learning, exploring, and applying knowledge, so as to apply what they have learned and use it to bring learning.
3. Teachers design in-depth learning topics and guide students to carry out research, which can also better mobilize students' enthusiasm for learning and applying knowledge.
It can be said that the ultimate goal of our teaching is for students to learn knowledge, apply knowledge, form abilities, and become students' own development skills. Therefore, we allow students to turn knowledge into knowledge and skills that can be seen, imagined, and used.
In this way, we will choose the appropriate entry point for teaching and guide students to carry out the journey of knowledge application and exploration, so that students' learning momentum can be stimulated, their interest can be persisted, and all difficulties will become easier and become easier. Freely, they no longer regard learning and applying knowledge as a painful thing.
The best direction for teachers to design topics is: visible, accessible, and usable; again, the above standards can be achieved with the help of instruments; the worst thing is, the above standards can be achieved with the help of the Internet.
In this way, most of the students can carry out in-depth learning, and at the same time, they can achieve the optimal level. The above points are my understanding of deep learning thinking and work development. If it is inappropriate, I hope that all leaders and colleagues will correct me.
Talking about how to make deep learning in junior high school mathematics classroom more effective
Deep learning is relative to shallow learning.
The main characteristics of in-depth learning in junior high school mathematics are active understanding and critical acceptance, activation of experience and construction of new knowledge, knowledge integration and deep processing, grasping the essence and penetration of ideas, effective migration and problem solving. Therefore, the promotion strategies for in-depth learning in junior high school mathematics are proposed: Creation of situations; problem-driven; knowledge integration; cooperative inquiry; incentive evaluation 1. Characteristics of deep learning in junior high school mathematics At present, there are few studies on deep learning in junior high school mathematics.
Combining the reality of junior high school mathematics teaching and the research results of deep learning by domestic and foreign scholars, we believe that junior high school mathematics deep learning is relative to the passive, isolated and mechanical shallow learning that appears in junior high school mathematics teaching. On the basis of learning, transform from receptive learning to inquiry learning, develop from low-order thinking ability to high-order thinking ability, extend from simple intuitive knowledge structure to extended abstract knowledge structure, and realize active learning based on original knowledge and experience. The process of constructing, gradually improving the personal mathematical knowledge system, and effectively transferring and applying it to real situations.
From this, the main features that in-depth learning of mathematics in junior high school should be further obtained: (1) Active understanding and critical acceptance of in-depth learning of mathematics in junior high school should be based on the understanding of existing mathematical knowledge, and maintain a critical or skeptical attitude towards new mathematical knowledge attitude, and incorporate it into the original cognitive structure; through questioning and discrimination (rather than blindly adapting and accepting), the understanding of mathematical knowledge should be deepened, and the awareness of active learning and the ability of deep thinking should be improved.
(2) Activating experience and constructing new knowledge Eric Jensen and LeAnn Nickelsen proposed in "7 Powerful Strategies for Deep Learning": "Every student has a different schema or background knowledge when embarking on a learning journey.
"In-depth learning of mathematics in junior high school needs to activate existing experience, through the interaction of new and old mathematical knowledge, realize the assimilation and adaptation of knowledge, form an understanding of mathematical knowledge, and construct new knowledge.
(3) Knowledge integration and deep processing Nelson Laird and others, through theoretical analysis and empirical research on the deep learning scale developed by scholars such as Biggs, Entwistle and Ramsden, found that deep learning can be deconstructed into high-level learning, integrated learning, and reflective learning. Study these three interrelated parts.
Mathematical knowledge does not exist in isolation, there are inextricably linked between them.
In the in-depth study of junior middle school mathematics, students need to follow this rule, straighten out the corresponding relationship, establish the connection between old and new knowledge and information, and integrate them together through deep processing, making it the key to solving mathematical problems and developing thinking ability Strategy.
(4) Grasping the essence and infiltrating thought Mathematical knowledge may be forgotten, but mathematical thought will accompany a person's life. Through mathematical thinking, the essence of mathematics can be revealed.
Therefore, in-depth learning of mathematics in junior high schools requires students to flexibly use mathematical thinking, deeply grasp the essence of mathematics, and improve their personal thinking quality and learning efficiency.
(5) Effective transfer and problem solving Effective transfer and problem solving are the core features of deep learning. Students are required to activate existing experience, draw inferences in similar situations, and critically understand, transfer and apply in new situations.
In this regard, students can gradually improve their original knowledge and experience on the basis of superficial learning, actively construct a personal mathematical knowledge system, and effectively transfer and apply it to real situations.
2. Promotion strategies for in-depth learning of mathematics in junior high schools (1) Creating situations The process of in-depth learning of mathematics in junior high schools is a process in which students actively construct new knowledge. Absorb nutrients naturally and acquire new knowledge actively.
That is, context is one of the elements that facilitate deep learning. Therefore, teachers should not instill mathematical knowledge directly into students, but should allow students to experience the process of questioning, inquiry, induction, and generalization by creating situations to generate new knowledge independently. (2) Question-driven Questions are the starting point and driving force of thinking.
Problem-centered teaching is to guide students to solve problems, help students master knowledge, develop abilities, and gradually develop a good way of thinking.
Teachers should mobilize students' enthusiasm for thinking and exploration by setting hierarchical and flexible questions, and guide students' understanding from shallow to deep, from the surface to the inside; they should also stimulate students to ask valuable questions themselves to improve their higher-order thinking ability .
(3) Knowledge Integration Knowledge is not isolated, but like every node on a large network, there are innumerable connections between them. Students should also follow this point in learning, be good at discovering the connection between knowledge, integrate new knowledge with previously learned knowledge, and make it a part of the existing knowledge structure.
Deep learning especially emphasizes knowledge integration, focuses on critical understanding, promotes transfer and application, and is problem-solving-oriented. Therefore, it emphasizes multi-dimensional knowledge integration and new knowledge construction for non-well-structured problem solving in complex situations.
(4) The practice of cooperative inquiry has proved that the learning method of cooperative inquiry is more likely to stimulate students' interest in learning mathematics and broaden the breadth of students' participation in classroom activities; only by guiding students to actively explore, communicate with each other, communicate with each other, inspire each other, and complement each other. Students have a deep understanding and flexible application of mathematical knowledge.
This is fundamentally different from the traditional teaching that only focuses on teachers’ teaching and students’ passive acceptance. Teachers are required to respect students’ individuality, give play to students’ subjectivity, and guide students to actively discover, ask, and analyze problems through peer assistance. , problem solving, and deep learning.
(5) Pay attention to multiple evaluations and strengthen learning motivation. Teaching evaluation is an important means of teaching regulation. It is the link between teachers and students' thinking and emotions, and directly affects students' psychological activities. In the mathematics classroom of deep learning, the evaluation of students should not be limited to the scores and the level of knowledge and skill application.
Instead, more attention should be paid to whether students' mathematical emotions, attitudes, and values have been effectively developed, and whether students' changes and developments in the learning process. Evaluation methods need to be diversified. 1. Observation and evaluation method.
The classroom is dynamic. Classroom observation and evaluation can help teachers better understand the problems existing in the classroom, and allow teachers to more clearly grasp the learning situation of students, so as to improve teaching methods, promote effective learning of students and accelerate their own development.
For example, when students are actively performing or thinking sparks, teachers should give appropriate encouragement and positive evaluation in time, so that students can experience the joy of success, so as to achieve greater success in success.
In the classroom, teachers can ask students questions, understand the students' understanding of the knowledge they have learned according to the students' answers, adjust the subject teaching in a timely manner, and improve students' enthusiasm for learning mathematics by using motivating evaluations. 2. Test evaluation method.
For example, when the teacher finishes teaching a certain knowledge point, he can conduct a small test and let the students complete several knowledge-related questions. Through the completion of the students, it was found that most of the students had problems with the questions, and the teacher further analyzed and explained the reasons and types of mistakes. 3. Interview evaluation method.
Students can be convened for face-to-face communication, so as to understand students' views, thoughts and feelings, and then evaluate students.
Interview is a process of two-way communication between teachers and students. On the one hand, it can better grasp the individual differences of students and allow students to develop further. On the other hand, it can also let teachers understand their own teaching advantages and disadvantages.
Modern teaching philosophy emphasizes student-oriented, and the present is an era of pursuing the depth of thought. Junior high school mathematics teaching plays an important role in the growth process of students. It can not only cultivate students' meticulous thinking, but also cultivate students' mathematical rationality and harmony. See the depth of the problem.
Mathematics learning as a process of thought formation requires a certain depth, and it is inevitable to pursue deep learning in mathematics teaching in junior high schools.
How to combine deep learning with management to publish papers
First of all, the main question is what kind of basic knowledge is needed to understand the mathematical derivation of papers on deep learning theory. The answer to this question is actually very broad. Here, I assume that the subject of the question has mastered scores, advanced generations, probability theory, necessary statistical knowledge and the theoretical basis of Bayesian.
In addition, in recent years, some papers on ICLR and ICML have used more and more mathematical knowledge, including but not limited to real variables, functionals, point set topology, differential geometry, and abstract algebra.
Those with an engineering background often go up and see that people define a bunch of Greek letters and a bunch of fancy letters, and they will feel dizzy. If they want to supplement some mathematical foundations, they don't know where to start. My personal experience is that it is unrealistic and a waste of time for engineering students in the direction of deep learning to learn all these courses.
However, the following content is recommended even for engineering students: real variable function (this is the most frequently appearing part in the paper.
At least know what is a measurable set, what is an unmeasurable set, what is integrable, further Riemann integrable and Lebesgue integrable; understand the concept of lower measure) functional, variational method (this course is really It is difficult, and I only learned part of it, but for machine learning, the variational Euler-Lagrange equation must be known) Basic topology concepts (everyone likes to use the word manifold in current papers, as long as it comes to high-dimensional The data is manifold, and the source is here.
Another example is the proof of the perfect classifier in WGAN, which is actually a very basic proof of the metric space and Hausdorff space in the textbook) A little basic metric knowledge (there are also many in the paper, but I feel that I know the metric tensor and the index Mapping, knowing the geodesic equation is almost enough, and some deeper geometric concepts rarely appear) I have never learned deeper mathematics.
For example, this year's ICLR article on spherical CNN, I was also confused, but the above are basically enough for you to look at most in-depth theoretical articles from a high-level perspective.
What basic knowledge is required for deep learning?
Mathematical foundation If you can read and understand the mathematical formulas in deep learning papers smoothly, and can independently derive new methods, it means that you already have the necessary mathematical foundation.
Mastering the mathematical knowledge contained in the four mathematics courses of mathematical analysis, linear algebra, probability theory and convex optimization, and being familiar with the basic theories and methods of machine learning are the prerequisites for getting started with deep learning technology.
Because whether it is to understand the operation and gradient derivation of each layer in the deep network, or to formalize the problem or derive the loss function, it is inseparable from a solid foundation of mathematics and machine learning. Mathematical analysis In the advanced mathematics courses offered by engineering majors, the main learning content is calculus.
For general deep learning research and applications, it is necessary to focus on basic knowledge such as functions and limits, derivatives (especially composite function derivations), differentiation, integration, power series expansion, and differential equations. In the optimization process of deep learning, solving the first derivative of the function is the most basic work.
When it comes to the differential mean value theorem, Taylor's formula, and Lagrange multipliers, you shouldn't just feel familiar. Operations in linear algebra deep learning are often expressed as vector and matrix operations. Linear algebra is just such a branch of mathematics that takes vectors and matrices as its research objects.
What needs to be reviewed includes vectors, linear spaces, linear equations, matrices, matrix operations and their properties, and vector calculus.
When it comes to Jacobian and Hessian matrices, you need to know the exact mathematical form; when given a loss function in matrix form, you can easily solve for the gradient.
Probability theory Probability theory is a branch of mathematics that studies the law of random phenomena. Random variables have many applications in deep learning. Whether it is stochastic gradient descent, parameter initialization methods (such as Xavier), or Dropout regularization algorithms, they are inseparable from probability theory. theoretical support.
In addition to mastering the basic concepts of random phenomena (such as random experiments, sample space, probability, conditional probability, etc.), random variables and their distribution, it is also necessary to have a good understanding of the law of large numbers and central limit theorem, parameter estimation, hypothesis testing, etc. After understanding, you can further study a little bit of random process and Markov random chain.
Convex optimization Combining the above three basic mathematics courses, convex optimization can be said to be an applied course.
But for deep learning, since the commonly used deep learning optimization methods often only use first-order gradient information for stochastic gradient descent, practitioners do not actually need much "advanced" knowledge of convex optimization.
Understand the basic concepts of convex sets, convex functions, and convex optimization, master the general concepts of dual problems, master common unconstrained optimization methods such as gradient descent method, stochastic gradient descent method, and Newton method, and understand a little bit of equality-constrained optimization and inequality-constrained optimization method, which can meet the theoretical requirements for understanding optimization methods in deep learning.
In the final analysis of machine learning, deep learning is just a kind of machine learning method, and statistical machine learning is the de facto methodology in the field of machine learning.
Taking supervised learning as an example, you need to master representative machine learning techniques such as regression and classification of linear models, support vector machines and kernel methods, and random forest methods, and understand model selection and model reasoning, model regularization techniques, and model integration. , Bootstrap methods, probabilistic graphical models, etc.
Going a step further requires expertise in semi-supervised learning, unsupervised learning, and reinforcement learning.
Do deep learning papers have to disclose the code?
no. 1. Very unfriendly to the review. Some paper authors only give pseudocode, but it is very difficult to reproduce the results of the paper with pseudocode. Because for deep learning, every subtle parameter is very important, and a little difference may cause the result to be irreproducible.
And open code makes it easier to get your paper through review. Starting in 2019, ICML has added paper reproducibility as a factor for review. 2. It is unfair to researchers. Some researchers do not have access to large computing resources.
What if a large group publishes a paper and a graduate student needs to use the results? Is he expected to reproduce the research results of an engineering team of hundreds of people by himself? This is obviously unrealistic.
Open code allows researchers to keep up with the latest research results, which is crucial to maintaining the competitiveness of academia.