CN110288878A - Adaptive learning method and device - Google Patents
Adaptive learning method and device Download PDFInfo
- Publication number
- CN110288878A CN110288878A CN201910584394.2A CN201910584394A CN110288878A CN 110288878 A CN110288878 A CN 110288878A CN 201910584394 A CN201910584394 A CN 201910584394A CN 110288878 A CN110288878 A CN 110288878A
- Authority
- CN
- China
- Prior art keywords
- knowledge
- blocks
- learning
- student
- vector
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G09—EDUCATION; CRYPTOGRAPHY; DISPLAY; ADVERTISING; SEALS
- G09B—EDUCATIONAL OR DEMONSTRATION APPLIANCES; APPLIANCES FOR TEACHING, OR COMMUNICATING WITH, THE BLIND, DEAF OR MUTE; MODELS; PLANETARIA; GLOBES; MAPS; DIAGRAMS
- G09B19/00—Teaching not covered by other main groups of this subclass
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Physics & Mathematics (AREA)
- Educational Administration (AREA)
- Educational Technology (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The embodiment of the present invention provides a kind of adaptive learning method and device, belongs to machine learning techniques field.Include: the first blocks of knowledge currently learnt according to target learning path and student, determine candidate's blocks of knowledge set, includes all blocks of knowledge that student needs to learn in target learning path；According to the current learning states of student, determine probability when each blocks of knowledge is as object knowledge unit in candidate's blocks of knowledge set for optimal solution, and using the corresponding blocks of knowledge of maximum probability in candidate blocks of knowledge set as object knowledge unit, object knowledge unit is the next blocks of knowledge for needing to learn of student.Due to recommending next blocks of knowledge for needing to learn in combination with the learning state of the structure of knowledge and student, so as to accurately analyze student in the acquisition of knowledge degree of different moments, and recommendation results are made more to meet cognitive law, and then efficient learning path can be formulated personalizedly for different students.
Description
Technical field
The present invention relates to machine learning techniques field more particularly to a kind of adaptive learning methods and device.
Background technique
Traditional education at present, especially classroom instruction carry out general type education just for a class or a group, difficult
To meet student individuality demand.Meanwhile traditional education is big to educational resource demand, the present education inadequate resource the case where
Under, the phenomenon that being easy to produce the uneven situation of educational resource distribution, be easy to cause educational inequality.Now it is badly in need of a kind of adaptive
Learning method, to recommend the blocks of knowledge for being suitble to student's study to meet the individualized learning demand of different students.
Summary of the invention
To solve the above-mentioned problems, the embodiment of the present invention provides one kind and overcomes the above problem or at least be partially solved
State the adaptive learning method and device of problem.
According to a first aspect of the embodiments of the present invention, a kind of adaptive learning method is provided, comprising:
The first blocks of knowledge currently learnt according to target learning path and student determines candidate's blocks of knowledge set, mesh
It include all blocks of knowledge that student needs to learn in mark learning path；
According to the current learning states of student, determine that each blocks of knowledge is as object knowledge in candidate's blocks of knowledge set
It is the probability of optimal solution when unit, and using the corresponding blocks of knowledge of maximum probability in candidate blocks of knowledge set as object knowledge
Unit, object knowledge unit are the next blocks of knowledge for needing to learn of student.
According to a second aspect of the embodiments of the present invention, a kind of adaptive learning device is provided, comprising:
First determining module, the first blocks of knowledge for currently being learnt according to target learning path and student are determined and are waited
Blocks of knowledge set is selected, includes all blocks of knowledge that student needs to learn in target learning path；
Second determining module, it is each in determining candidate's blocks of knowledge set to know for the current learning states according to student
When knowing unit as object knowledge unit it is the probability of optimal solution, and knows maximum probability is corresponding in candidate blocks of knowledge set
Unit is known as object knowledge unit, and object knowledge unit is the next blocks of knowledge for needing to learn of student.
According to a third aspect of the embodiments of the present invention, a kind of electronic equipment is provided, comprising:
At least one processor；And
At least one processor being connect with processor communication, in which:
Memory is stored with the program instruction that can be executed by processor, and the instruction of processor caller is able to carry out first party
Adaptive learning method provided by any possible implementation in the various possible implementations in face.
According to the fourth aspect of the invention, a kind of non-transient computer readable storage medium, non-transient computer are provided
Readable storage medium storing program for executing stores computer instruction, and computer instruction makes the various possible implementations of computer execution first aspect
In adaptive learning method provided by any possible implementation.
Adaptive learning method and device provided in an embodiment of the present invention, by current according to target learning path and student
First blocks of knowledge of study determines candidate's blocks of knowledge set.According to the current learning states of student, candidate's knowledge list is determined
Member set in each blocks of knowledge as object knowledge unit when be optimal solution probability, and by candidate blocks of knowledge set most
The corresponding blocks of knowledge of maximum probability is as object knowledge unit.Under recommending due to the learning state in combination with the structure of knowledge and student
One blocks of knowledge for needing to learn, so as to accurately analyze student in the acquisition of knowledge degree of different moments, and to push away
It recommends result and more meets cognitive law, and then efficient learning path can be formulated personalizedly for different students.
It should be understood that above general description and following detailed description be it is exemplary and explanatory, can not
Limit the embodiment of the present invention.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is this hair
Bright some embodiments for those of ordinary skill in the art without creative efforts, can be with root
Other attached drawings are obtained according to these attached drawings.
Fig. 1 is a kind of flow diagram of adaptive learning method provided in an embodiment of the present invention；
Fig. 2 is a kind of schematic diagram of target learning path provided in an embodiment of the present invention；
Fig. 3 is a kind of structural schematic diagram of preset model provided in an embodiment of the present invention；
Fig. 4 is a kind of structural schematic diagram of adaptive learning device provided in an embodiment of the present invention；
Fig. 5 is the block diagram of a kind of electronic equipment provided in an embodiment of the present invention.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art
Every other embodiment obtained without creative efforts, shall fall within the protection scope of the present invention.
Currently, adaptive learning method mainly has the following two kinds:
(1) based on the method for learning state
Since the learning ability of different students is different, hence for each student, learned brought by each blocks of knowledge
It is also different to practise income.Therefore, the learning state and its changing rule of different students is different.It is managed based on Item Response Pattern
By can speculate the state and ability of student, and recommend to student based on this according to performance of the student on different blocks of knowledge
The moderate blocks of knowledge of difficulty.In addition to this it is possible to adaptive learning process is considered as a Markovian decision process, it will
The transfer matrix of the evolution process of student's learning state Markovian decision process is dug to simulate using nitrification enhancement
Dig the relationship between learning state and blocks of knowledge.
(2) method of knowledge based structure
By carrying out data analysis to the relationship between blocks of knowledge, carried out in conjunction with similitude, difficulty of blocks of knowledge etc.
Recommend.Specifically, knowledge mapping can be introduced wherein, by the way that knowledge hierarchy is expressed as map form, knowledge based unit
Feature and relationship formulate recommendation rules to plan learning path for student.Alternatively, the learning ability due to student can reflect
It learns on track, and similar learning ability corresponds to the similar structure of knowledge, so as to use the methods of collaborative filtering to
Student recommends similar blocks of knowledge.For example, the method migration that traditional electric business is recommended is applied in education recommendation.
For above-mentioned first way, the method based on student's learning state can not effectively utilize existing knowledge knot
Structure may provide the illogical learning path for violating human cognitive rule.And it is directed to the above-mentioned second way, based on knowing
Know structure method can not effectively for different students formulate individualized learning method, can only as traditional education method, from
Group's level cooks up the learning path of general type, to not can guarantee the high efficiency of study.
The problem of for above two mode, the embodiment of the invention provides a kind of adaptive learning methods.It should
Method can be used for the recommendation scene of blocks of knowledge, and the present invention is not especially limit this.Specifically, since student exists
When learning certain subject, different blocks of knowledge is front and back place mat, such as finishes function and the limit, could learn derivative and micro- product
Point, so that student after having learnt a certain blocks of knowledge, needs to recommend next blocks of knowledge for needing to learn to it.Referring to figure
1, this method comprises:
101, the first blocks of knowledge currently learnt according to target learning path and student determines candidate's blocks of knowledge collection
It closes, includes all blocks of knowledge that student needs to learn in target learning path.
Wherein, target learning path refers to universality learning path, may include blocks of knowledge and blocks of knowledge it
Between sequencing, can be specifically indicated by way of the connection schematic diagram with direction.First blocks of knowledge can root
It is determined according to the studying progress of student, as student is currently learning the 3rd blocks of knowledge or the 3rd blocks of knowledge just
Study finishes, then the 3rd blocks of knowledge is the first blocks of knowledge that the student currently learns.In addition, in target learning path
Comprising first blocks of knowledge, candidate blocks of knowledge set is also the blocks of knowledge screened from target learning path.
102, according to the current learning states of student, determine that each blocks of knowledge is as target in candidate's blocks of knowledge set
It is the probability of optimal solution when blocks of knowledge, and using the corresponding blocks of knowledge of maximum probability in candidate blocks of knowledge set as target
Blocks of knowledge, object knowledge unit are the next blocks of knowledge for needing to learn of student.
Wherein, the current learning states of student may include the history performance of the test of the student and the learning objective of the student,
And can be embodied by way of vector, the present invention is not especially limit this.It should be noted that according to upper
After the mode of stating determines next blocks of knowledge for needing to learn, the blocks of knowledge can be recommended to student.If the student is learning
The blocks of knowledge learns the blocks of knowledge that is over, then the first knowledge that can currently learn the blocks of knowledge as the student
Unit, and continue as the student according to above-mentioned steps 101 to 102 and recommend next blocks of knowledge.By above-mentioned recommendation process, directly
To student's study, the blocks of knowledge that each step is recommended can form a learning path.The learning path can be used as
New target learning path, and it is used for the process of adaptive learning.
Method provided in an embodiment of the present invention passes through the first knowledge list currently learnt according to target learning path and student
Member determines candidate's blocks of knowledge set.According to the current learning states of student, each knowledge in candidate's blocks of knowledge set is determined
It is the probability of optimal solution when unit is as object knowledge unit, and by the corresponding knowledge of maximum probability in candidate blocks of knowledge set
Unit is as object knowledge unit.Since the learning state in combination with the structure of knowledge and student is recommended next to need what is learnt to know
Know unit, so as to accurately analyze student in the acquisition of knowledge degree of different moments, and recommendation results is made more to meet cognition
Rule, and then can efficient learning path be formulated for different students personalizedly.
Content based on the above embodiment is determined as a kind of alternative embodiment in the current learning states according to student
When each blocks of knowledge is as object knowledge unit in candidate blocks of knowledge set for the probability of optimal solution before, can also obtain
The current learning states of student.The embodiment of the present invention does not limit the mode for the current learning states for obtaining student specifically, wraps
It includes but is not limited to: being test and recorded according to the history of student, obtain the current learning states vector of student, history test record is used for
Indicate the test result to blocks of knowledge in target learning path；The instruction vector for obtaining student, by instruction vector and current
Current learning states of the vector obtained after state vector is spliced as student are practised, instruction vector is for indicating that target learns
As the blocks of knowledge of learning objective in path.
Wherein, history test record refers to the blocks of knowledge for examining in the test of each history, and student knows these
Know the answer situation or study situation of unit.Each history test record can test vector by a history and carry out table
Show, all history test records can be indicated x=(x by following sequence1, x2...).With x1For, x1Indicate be
First time history test record answer or study situation namely first time history test record corresponding history test to
Amount, x2What is indicated is that second of history test records corresponding history test vector, and subsequent vector is similarly.It is with answer situation
Example, x1Dimension can be twice of blocks of knowledge quantity.
For example, if the blocks of knowledge that only an examination ID is 130 in the test of first time history, and has examined a problem, and be somebody's turn to do
Student has answered questions the corresponding topic of blocks of knowledge that ID is 130, then x1=(0,0 ..., 0,1(261), 0,0 ..., 0).Wherein,
" 261 " refer to x1In the 261st dimension element, it is corresponding that the value of the element is that 1 expression student has answered questions blocks of knowledge that ID is 130
Topic.The corresponding topic of blocks of knowledge that ID is 130 if the student has answered wrong, x1=(0,0 ..., 0,1(260), 0,
0 ..., 0).Wherein, " 260 " refer to x1In the 260th dimension element, the value of the element is that have answered ID wrong be 130 to 1 expression student
The corresponding topic of blocks of knowledge.
That is, mistake is answered questions or answered to the corresponding topic of each blocks of knowledge, the element that can pass through two dimensions carries out table
Show.For example, ID be 1 the corresponding topic of blocks of knowledge answer questions or answer mistake, can by the 1st dimension and the 2nd dimension element into
Row indicates.Mistake is answered questions or answered to the corresponding topic of blocks of knowledge that ID is 130, can pass through the 260th dimension and the member of the 261st dimension
Element is indicated.
As shown in the above, history test record can test vector by history and be indicated, and these are gone through
History tests the sequence x=(x of vector composition1, x2...) and it can reflect the learning state of student and the differentiation of student study situation
Rule.Therefore, vector is test according to the history of student, can further obtains the current learning states vector of student.Wherein, when
Preceding learning state vector can reflect the learning state of the student after multiple history test record, can also reflect that student learns
The development law of situation.
In addition, the dimension of instruction vector can be identical as blocks of knowledge quantity.By the way that all blocks of knowledge are numbered,
Each blocks of knowledge can correspond to the element of a dimension.For example, i-th of blocks of knowledge can correspond to i-th dimension in instruction vector
Element.I-th of blocks of knowledge, which is as, to be indicated if the element of i-th dimension is 1 for the element of i-th dimension in instruction vector
Practise the blocks of knowledge of target.If the element of i-th dimension is 0, i-th of blocks of knowledge can be indicated not as the knowledge of learning objective
Unit.It certainly, can also in turn during actual implementation namely 1 represents blocks of knowledge not as learning objective, 0 generation
Table is the blocks of knowledge as learning objective, and the present invention is not especially limit this.
It should be noted that the blocks of knowledge as learning objective can not make this with more than one, the embodiment of the present invention
It is specific to limit.For example, instruction vector can be (0,0,0,0,1(5),0,0,1(8),0,...,0,1(100),0,…,0).Wherein,
" 5 ", " 8 " and " 100 " indicate that the 5th, the 8th and the 100th blocks of knowledge is the blocks of knowledge as learning objective.It is obtaining
After the current learning states vector and instruction vector of student, the two can be spliced, thus the vector that will be obtained after splicing
Current learning states as student.
Method provided in an embodiment of the present invention is recorded by being test according to the history of student, obtains the current study of student
State vector.The instruction vector for obtaining student, the vector that will indicate that vector and current learning states vector obtain after being spliced
Current learning states as student.Record junior scholar is test by multiple history since the current learning states of student can reflect
Raw learning state can also reflect that student learns the development law of situation, thus the subsequent current learning states according to student,
Can efficient learning path be formulated for different students personalizedly.
Content based on the above embodiment, as a kind of alternative embodiment, history test is recorded as history test vector；Phase
Ying Di, the embodiment of the present invention do not test record to according to the history of student, obtain the mode of the current learning states vector of student
Make it is specific limit, including but not limited to: each history test vector being input in preset model, the output test moment is the latest
History tests the corresponding learning state vector of vector, and as current learning states vector.
Specifically, each history test vector can correspond to a learning state vector.For example, history tests vector x1
Learning state vector S can be corresponded to1, history test vector xtLearning state vector S can be corresponded tot.Wherein, preset model can be used for
The answer or study situation of the student is test in prediction next time.The input of preset model can test vector for different history, defeated
Out can be to test the answer of the student next time or learning situation, output result can also be indicated by vector.Due to going through every time
History test was carried out in different moments, so that history test vector has sequence in time.And test the moment the latest
History tests the corresponding learning state vector of vector, and it is to be able to reflect that all history, which tests vector, before being combined due to it
The current learning states of the student and study situation development law, so as to as current learning states vector.In addition, default
Model can be specially shot and long term memory models, can also chase after for the deep knowledge after being improved based on shot and long term memory models
Track model, the present invention is not especially limit this.
Method provided in an embodiment of the present invention, since current learning states vector is the current study for being able to reflect the student
State and study situation development law can be not classmate thus subsequent current learning states according to student personalizedly
It is raw to formulate efficient learning path.
Content based on the above embodiment, as a kind of alternative embodiment, preset model include at least embeding layer, hidden layer and
Full articulamentum；Correspondingly, the embodiment of the present invention is not input in preset model to by each history test vector, exports each go through
The mode of the corresponding learning state vector of history test vector specifically limits, including but not limited to: each history is test vector
It is input to embeding layer, exports the corresponding study characterization vector of each history test vector；Each study characterization vector is input to
Hidden layer exports the corresponding hidden vector of learning state of each history test vector；By the hidden vector of initial learning state and each study
The hidden vector of state is input to full articulamentum, the corresponding learning state vector of history test vector of output test moment the latest.
Specifically, since history test vector may be more sparse, to sparse vector can be become by embeding layer
For dense vector, to be used to compress the characterization of study or answer situation.Wherein, the structure of preset model and output learning state to
The process of amount can refer to Fig. 2.In Fig. 2, x1To xtIndicate that history tests vector, x1' to xt' indicate to export after embeding layer
Study characterize vector, h1To htIndicate the hidden vector of learning state exported after hidden layer, h0Indicate initial learning state it is hidden to
Amount, S1To StIndicate the learning state vector exported after full articulamentum.As shown in Figure 2, S1It is based on h0And h1It obtains, S2
It is based on h0、h1And h2It obtains, subsequent and so on, StIt is based on h0To htIt obtains.
It is calculated it should be noted that the initial hidden vector of learning state mainly plays the role of auxiliary, it is possible
Error can be with from S1It calculates to StAnd gradually weaken.In addition, such as Fig. 2, preset model can actually export each history test to
Corresponding learning state vector is measured, the embodiment of the present invention is mainly needed using St, namely test moment history the latest test to
Corresponding learning state vector is measured, and as current learning states vector.
Method provided in an embodiment of the present invention, since current learning states vector is the current study for being able to reflect the student
State and study situation development law can be not classmate thus subsequent current learning states according to student personalizedly
It is raw to formulate efficient learning path.
Content based on the above embodiment, as a kind of alternative embodiment, the embodiment of the present invention does not learn to according to target
The first blocks of knowledge that path and student currently learn determines that the mode of candidate's blocks of knowledge set specifically limits, including but
It is not limited to: determining the second blocks of knowledge and target learning path in target learning path before the first blocks of knowledge in m jump
In after the first blocks of knowledge n jump in third blocks of knowledge, m and n are the positive integer not less than 1；According to the first knowledge list
Blocks of knowledge in member, the second blocks of knowledge, third blocks of knowledge and target learning path as learning objective, determines that candidate knows
Know unit set.
Wherein, the value of m and n can be configured according to demand, and the two may be the same or different, and the present invention is implemented
Example is not especially limited this.As shown in figure 3, Fig. 3 is a kind of schematic diagram of target learning path.In Fig. 3, each node
A blocks of knowledge is represented, different blocks of knowledge is distinguished by the label in node.With the node marked as 3 for the
One blocks of knowledge, for m is 1, then before the first blocks of knowledge 1 jump in the second blocks of knowledge, the as corresponding section of label 1
Point.By taking n is 2 as an example, after the first blocks of knowledge 2 jump in third blocks of knowledge, for node marked as 4 and marked as 8.?
After determining the second blocks of knowledge and third blocks of knowledge in target learning path, can directly it be known by the second blocks of knowledge and third
Know unit and forms candidate blocks of knowledge set.It should be noted that the respective quantity of the second blocks of knowledge and third blocks of knowledge
It may more than one.
It should also be noted that, according to conventional mode of learning, student after having learnt the first blocks of knowledge marked as 3,
It should need to continue to learn backward.But it is considered that student learns a possibility that also review in need, to be located at the first knowledge list
The second blocks of knowledge before member be also included in it is subsequent may need within the scope of the considerations of learning, and then the second blocks of knowledge is also put
It is placed in candidate blocks of knowledge set.
Method provided in an embodiment of the present invention, since the second blocks of knowledge before being located at the first blocks of knowledge is also included in
It is subsequent to need within the scope of the considerations of learning, it is reviewed so as to allow student to realize, to reach better learning effect.
In view of student's learning process need to be in accordance with target learning path and to learn terminal (namely knowing as learning objective
Know unit) for the purpose of, and if directly forming candidate blocks of knowledge set by the second blocks of knowledge and third blocks of knowledge, it can
It can will lead in candidate blocks of knowledge set that there are blocks of knowledge cannot reach study terminal, so that students'learning is not
Meet cognitive law.For the situation, content based on the above embodiment, as a kind of alternative embodiment, the embodiment of the present invention
Not to according in the first blocks of knowledge, the second blocks of knowledge, third blocks of knowledge and target learning path as learning objective
Blocks of knowledge determines that the mode of candidate's blocks of knowledge set specifically limits, including but not limited to: being based on preset condition and target
Learning path screens the second blocks of knowledge and third blocks of knowledge, and forms candidate by the blocks of knowledge after screening and know
Know unit set；Wherein, preset condition is for energy and the first blocks of knowledge and as composition connection road between the blocks of knowledge of target
Diameter.
For example, in Fig. 3, node 1,2,3,4,8,9 can make up communication path, and node 2,0,1,3, due to section
Line directionality problem between point, can not form communication path.
Method provided in an embodiment of the present invention is obtained by screening to the second blocks of knowledge and third blocks of knowledge
Candidate blocks of knowledge set, it is subsequent to determine based on the obtained candidate blocks of knowledge set of screening and next need what is learnt to know
Know unit.Due to each blocks of knowledge and the first blocks of knowledge in candidate blocks of knowledge set and as the blocks of knowledge of target
Between can form communication path, thus make based on this determine learning path meet cognitive law.
Content based on the above embodiment, as a kind of alternative embodiment, the embodiment of the present invention is not worked as to according to student
Preceding learning state determines probability when each blocks of knowledge is as object knowledge unit in candidate's blocks of knowledge set for optimal solution
Mode specifically limit, including but not limited to: obtain from the first blocks of knowledge learn into target learning path as study
Generated learning ability incremental value after the blocks of knowledge of target；According to learning ability incremental value, determine in strategy network model
The final value of parameter preset, and current learning states are input to tactful network model, it exports in candidate blocks of knowledge set
Each blocks of knowledge is the probability of optimal solution when as object knowledge unit.
Wherein, the above process can be realized by way of intensified learning, namely pass through tactful network model and value network
Network model determines the next blocks of knowledge for needing to learn of student.The process of intensified learning is mainly concerned with following three element,
Respectively state, movement and reward.
Wherein, state is the current learning states for referring to student.For the step 101 and step in above-described embodiment
102, determine that the blocks of knowledge that the next needs of student learn is a movement each time.In addition, in learning process, reward
Signal is always 0, until study to the blocks of knowledge namely learning process as target terminates.It, can after learning process
To define the learning ability incremental value of student as prize signal.Wherein, learning ability incremental valueCalculating process can refer to
Following formula:
In above-mentioned formula,Indicate learning ability incremental value, EsIndicate the performance of the test when study stage starts, EeTable
Performance of the test after the dendrography habit stage, EsupIndicate the full marks value of test.Wherein it is possible to before the study stage starts and tie
Shu Hou makes student respectively participate in primary test, to respectively obtain E according to test resultsAnd Ee。
Based on above content, the optimum target of intensified learning can be provided with mathematical form:
In above-mentioned formula, γ is the discount factor constant in intensified learning, and N is the total degree in the study stage, rjTable
Show the prize signal learnt every time, RiIndicate to move that (namely what is indicated is student from the first knowledge list to the movement of N step from the i-th step
Meta learning is to the blocks of knowledge for being used as target) between each step movement the sum of prize signal.
Content based on the above embodiment, as a kind of alternative embodiment, the embodiment of the present invention is not to according to learning ability
Incremental value determines that the mode of the final value of parameter preset in strategy network model specifically limits, including but not limited to: will work as
Preceding learning state is input to PN model, the value of parameter preset in PN model is adjusted, so that value network
Difference between the output result and learning ability incremental value of model is minimum, and using the value of parameter preset when difference minimum as
The final value of parameter preset, PN model and tactful network model include parameter preset.
Specifically, can first define a PN model v (| θv), for estimating that a certain state can obtain in future
The reward income total value v obtainedi=v (statei|θv).Wherein, θvRefer to the parameter preset in PN model.It will be random
Strategy is applied in PN model, performer-judge man algorithm can be applied on the recommendation generation of blocks of knowledge.?
When proceeding to the movement of the i-th step, Policy-Gradient function can refer to following formula:
In above-mentioned formula, π () indicates the strategic function in tactful network model.It is every in candidate blocks of knowledge set
One blocks of knowledge is equivalent to a kind of strategy of learning path, in given current learning states stateiWith the value of parameter preset
Afterwards, when each blocks of knowledge in exportable candidate blocks of knowledge set is as next blocks of knowledge for needing to learn, make
For the probability value of optimal movement.
Wherein, the final value of parameter preset can be determined by the loss function of PN model, loss function
It can refer to following formula:
The loss function of the Policy-Gradient function of tactful network model and PN model is combined, it is available
The loss function of whole network, with specific reference to following formula:
In above-mentioned formula, α and β are hyper parameter.It, can be with by adjusting the value of parameter preset in PN model
So that the output result v of PN modeliIt changes, until the output result and learning ability increment of PN model
Difference between value is minimum, can be using value at this time as parameter preset θvFinal value.Due to tactful network model with
PN model includes parameter preset, and after determining parameter preset, current learning states can be input to tactful network
Model, i.e., each blocks of knowledge is the general of optimal solution when as object knowledge unit in exportable candidate blocks of knowledge set
Rate, so as to using the corresponding blocks of knowledge of maximum probability as the next blocks of knowledge for needing to learn of student.It needs to illustrate
It is that optimal solution refers to can reach the optimum target of intensified learning, namely makes the output result of PN model and learn
Difference between habit ability incremental value is minimum.
Method provided in an embodiment of the present invention, by converting Markovian decision gradually for learning path recommendation problem
Problem, and performer-reviewer's algorithm is applied, dynamic updates Generalization bounds, to sequentially recommend to be able to achieve height to different students
Imitate the blocks of knowledge of study.
Content based on the above embodiment, the embodiment of the invention provides a kind of adaptive learning device, adaptive
Device is practised for executing the adaptive learning method provided in above method embodiment.Referring to fig. 4, which includes:
First determining module 401, the first blocks of knowledge for currently being learnt according to target learning path and student determine
Candidate blocks of knowledge set includes all blocks of knowledge that student needs to learn in target learning path；
Second determining module 402 determines each in candidate's blocks of knowledge set for the current learning states according to student
It is the probability of optimal solution when blocks of knowledge is as object knowledge unit, and maximum probability in candidate blocks of knowledge set is corresponding
For blocks of knowledge as object knowledge unit, object knowledge unit is the next blocks of knowledge for needing to learn of student.
As a kind of alternative embodiment, the device further include:
First obtains module, records for being test according to the history of student, obtains the current learning states vector of student, go through
History test record is for indicating the test result to blocks of knowledge in target learning path；
Second obtains module, for obtaining the instruction vector of student；
Splicing module, for the vector obtained after vector and current learning states vector are spliced will to be indicated as student
Current learning states, instruction vector is used to indicate blocks of knowledge in target learning path as learning objective.
Module is obtained as a kind of alternative embodiment, first, for each history test vector to be input to preset model
In, the corresponding learning state vector of history test vector of output test moment the latest, and as current learning states vector.
As a kind of alternative embodiment, preset model includes at least embeding layer, hidden layer and full articulamentum；Correspondingly, first
Module is obtained, for each history test vector to be input to embeding layer, exports the corresponding learning table of each history test vector
Levy vector；Each study characterization vector is input to hidden layer, exports the corresponding hidden vector of learning state of each history test vector；
The hidden vector of initial learning state and the hidden vector of each learning state are input to full articulamentum, the history of output test moment the latest
Test the corresponding learning state vector of vector.
As a kind of alternative embodiment, the first determining module 401, for determining the first blocks of knowledge in target learning path
Third knowledge list in the second blocks of knowledge and target learning path in m jump before after the first blocks of knowledge in n jump
Member, m and n are the positive integer not less than 1；According to the first blocks of knowledge, the second blocks of knowledge, third blocks of knowledge and target
The blocks of knowledge in path as learning objective is practised, determines candidate's blocks of knowledge set.
As a kind of alternative embodiment, the second determining module 402, comprising:
Acquiring unit, for obtaining the knowledge learnt into target learning path as learning objective from the first blocks of knowledge
Generated learning ability incremental value after unit；
Second determination unit is used for according to learning ability incremental value, and parameter preset is final in determining strategy network model
Value；
Second output unit exports candidate blocks of knowledge collection for current learning states to be input to tactful network model
Each blocks of knowledge is the probability of optimal solution when as object knowledge unit in conjunction.
As a kind of alternative embodiment, the second determination unit, for current learning states to be input to PN model,
The value of parameter preset in PN model is adjusted, so that the output result of PN model and learning ability incremental value
Between difference it is minimum, and using the value of parameter preset when difference minimum as the final value of parameter preset, value network mould
Type and tactful network model include parameter preset.
Device provided in an embodiment of the present invention passes through the first knowledge list currently learnt according to target learning path and student
Member determines candidate's blocks of knowledge set.According to the current learning states of student, each knowledge in candidate's blocks of knowledge set is determined
It is the probability of optimal solution when unit is as object knowledge unit, and by the corresponding knowledge of maximum probability in candidate blocks of knowledge set
Unit is as object knowledge unit.Since the learning state in combination with the structure of knowledge and student is recommended next to need what is learnt to know
Know unit, so as to accurately analyze student in the acquisition of knowledge degree of different moments, and recommendation results is made more to meet cognition
Rule, and then can efficient learning path be formulated for different students personalizedly.
Fig. 5 illustrates the entity structure schematic diagram of a kind of electronic equipment, as shown in figure 5, the electronic equipment may include: place
Manage device (processor) 510, communication interface (Communications Interface) 520,530 He of memory (memory)
Communication bus 540, wherein processor 510, communication interface 520, memory 530 complete mutual lead to by communication bus 540
Letter.Processor 510 can call the logical order in memory 530, to execute following method: according to target learning path and
Raw the first blocks of knowledge currently learnt, determines candidate's blocks of knowledge set, includes that student needs to learn in target learning path
All blocks of knowledge；According to the current learning states of student, each blocks of knowledge conduct in candidate's blocks of knowledge set is determined
Be the probability of optimal solution when object knowledge unit, and using the corresponding blocks of knowledge of maximum probability in candidate blocks of knowledge set as
Object knowledge unit, object knowledge unit are the next blocks of knowledge for needing to learn of student.
In addition, the logical order in above-mentioned memory 530 can be realized by way of SFU software functional unit and conduct
Independent product when selling or using, can store in a computer readable storage medium.Based on this understanding, originally
Substantially the part of the part that contributes to existing technology or the technical solution can be in other words for the technical solution of invention
The form of software product embodies, which is stored in a storage medium, including some instructions to
So that a computer equipment (can be personal computer, electronic equipment or the network equipment etc.) executes each reality of the present invention
Apply all or part of the steps of a method.And storage medium above-mentioned include: USB flash disk, mobile hard disk, read-only memory (ROM,
Read-Only Memory), random access memory (RAM, Random AccessMemory), magnetic or disk etc. it is various
It can store the medium of program code.
The embodiment of the present invention also provides a kind of non-transient computer readable storage medium, is stored thereon with computer program,
The computer program is implemented to carry out the various embodiments described above offer method when being executed by processor, for example, according to target
The first blocks of knowledge that learning path and student currently learn determines candidate's blocks of knowledge set, includes in target learning path
Student needs all blocks of knowledge learnt；According to the current learning states of student, determine each in candidate's blocks of knowledge set
It is the probability of optimal solution when blocks of knowledge is as object knowledge unit, and maximum probability in candidate blocks of knowledge set is corresponding
For blocks of knowledge as object knowledge unit, object knowledge unit is the next blocks of knowledge for needing to learn of student.
The apparatus embodiments described above are merely exemplary, wherein unit can be as illustrated by the separation member
Or may not be and be physically separated, component shown as a unit may or may not be physical unit, i.e.,
It can be located in one place, or may be distributed over multiple network units.It can select according to the actual needs therein
Some or all of the modules achieves the purpose of the solution of this embodiment.Those of ordinary skill in the art are not paying creative labor
In the case where dynamic, it can understand and implement.
Through the above description of the embodiments, those skilled in the art can be understood that each embodiment can
It realizes by means of software and necessary general hardware platform, naturally it is also possible to pass through hardware.Based on this understanding, on
Stating technical solution, substantially the part that contributes to existing technology can be embodied in the form of software products in other words, should
Computer software product may be stored in a computer readable storage medium, such as ROM/RAM, magnetic disk, CD, including several fingers
It enables and using so that a computer equipment (can be personal computer, server or the network equipment etc.) executes each implementation
Method described in certain parts of example or embodiment.
Finally, it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations；Although
Present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: it still may be used
To modify the technical solutions described in the foregoing embodiments or equivalent replacement of some of the technical features；
And these are modified or replaceed, technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution spirit and
Range.
Claims (10)
1. a kind of adaptive learning method characterized by comprising
The first blocks of knowledge currently learnt according to target learning path and student determines candidate's blocks of knowledge set, the mesh
It include all blocks of knowledge that the student needs to learn in mark learning path；
According to the current learning states of the student, determine that each blocks of knowledge is as target in the candidate blocks of knowledge set
Be the probability of optimal solution when blocks of knowledge, and using the corresponding blocks of knowledge of maximum probability in the candidate blocks of knowledge set as
The object knowledge unit, the object knowledge unit are the next blocks of knowledge for needing to learn of the student.
2. adaptive learning method according to claim 1, which is characterized in that the current study according to the student
State, determine when each blocks of knowledge is as object knowledge unit in the candidate blocks of knowledge set for the probability of optimal solution it
Before, further includes:
It is test and is recorded according to the history of the student, obtain the current learning states vector of the student, the history test note
Employ the test result in expression to blocks of knowledge in the target learning path；
The instruction vector for obtaining the student obtains after being spliced the instruction vector and the current learning states vector
Current learning states of the vector as the student, the instruction vector is for indicating in the target learning path as learning
Practise the blocks of knowledge of target.
3. adaptive learning method according to claim 2, which is characterized in that the history test is recorded as history test
Vector；Correspondingly, described test according to the history of the student records, and obtains the current learning states vector of the student, wraps
It includes:
Each history test vector is input in preset model, corresponding of history test vector of output test moment the latest
State vector is practised, and as the current learning states vector.
4. adaptive learning method according to claim 3, which is characterized in that the preset model includes at least insertion
Layer, hidden layer and full articulamentum；Correspondingly, described that each history test vector is input in preset model, the output test moment
The corresponding learning state vector of history test vector the latest, comprising:
Each history test vector is input to the embeding layer, export the corresponding study of each history test vector characterize to
Amount；
Each study characterization vector is input to the hidden layer, export the corresponding learning state of each history test vector it is hidden to
Amount；
The hidden vector of initial learning state and the hidden vector of each learning state are input to the full articulamentum, the output test moment is most
The corresponding learning state vector of history test vector in evening.
5. adaptive learning method according to claim 1, which is characterized in that described according to target learning path and student
The first blocks of knowledge currently learnt determines candidate's blocks of knowledge set, comprising:
Determine m before the first blocks of knowledge described in the target learning path jump in the second blocks of knowledge and the mesh
Third blocks of knowledge after first blocks of knowledge described in mark learning path in n jump, m and n are the positive integer not less than 1；
According to first blocks of knowledge, second blocks of knowledge, the third blocks of knowledge and the target learning path
The middle blocks of knowledge as learning objective determines the candidate blocks of knowledge set.
6. adaptive learning method according to claim 1, which is characterized in that the current study according to the student
State determines probability when each blocks of knowledge is as object knowledge unit in the candidate blocks of knowledge set for optimal solution,
Include:
Acquisition learns into the target learning path from first blocks of knowledge as institute after the blocks of knowledge of learning objective
The learning ability incremental value of generation；
According to the learning ability incremental value, the final value of parameter preset in strategy network model is determined, and will be described current
Learning state is input to the tactful network model, and each blocks of knowledge is as mesh in the output candidate blocks of knowledge set
It is the probability of optimal solution when mark blocks of knowledge.
7. adaptive learning method according to claim 6, which is characterized in that described according to the learning ability increment
Value determines the final value of parameter preset in strategy network model, comprising:
The current learning states are input to PN model, adjust parameter preset described in the PN model
Value, so that the difference between the output result of the PN model and the learning ability incremental value is minimum, and will
Final value of the value of the parameter preset as parameter preset when difference minimum, the PN model with it is described
Tactful network model includes the parameter preset.
8. a kind of adaptive learning device characterized by comprising
First determining module, the first blocks of knowledge for currently being learnt according to target learning path and student, determines that candidate knows
Know unit set, includes all blocks of knowledge that the student needs to learn in the target learning path；
Second determining module determines every in the candidate blocks of knowledge set for the current learning states according to the student
It is the probability of optimal solution when one blocks of knowledge is as object knowledge unit, and by maximum probability in the candidate blocks of knowledge set
Corresponding blocks of knowledge learns as the object knowledge unit, the next needs of the object knowledge unit student
Blocks of knowledge.
9. a kind of electronic equipment characterized by comprising
At least one processor；And
At least one processor being connect with the processor communication, in which:
The memory is stored with the program instruction that can be executed by the processor, and the processor calls described program to instruct energy
Enough methods executed as described in claim 1 to 7 is any.
10. a kind of non-transient computer readable storage medium, which is characterized in that the non-transient computer readable storage medium is deposited
Computer instruction is stored up, the computer instruction makes the computer execute the method as described in claim 1 to 7 is any.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910584394.2A CN110288878B (en) | 2019-07-01 | 2019-07-01 | Self-adaptive learning method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910584394.2A CN110288878B (en) | 2019-07-01 | 2019-07-01 | Self-adaptive learning method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110288878A true CN110288878A (en) | 2019-09-27 |
| CN110288878B CN110288878B (en) | 2021-10-08 |
Family
ID=68021364
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910584394.2A Active CN110288878B (en) | 2019-07-01 | 2019-07-01 | Self-adaptive learning method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110288878B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110928920A (en) * | 2019-11-19 | 2020-03-27 | 广东交通职业技术学院 | Knowledge recommendation method, system and storage medium based on improved position social contact |
| CN112053091A (en) * | 2020-09-28 | 2020-12-08 | 北京爱论答科技有限公司 | Data processing method and system based on learning operation |
| CN112180726A (en) * | 2020-09-29 | 2021-01-05 | 北京航空航天大学 | Spacecraft relative motion trajectory planning method based on meta-learning |
| CN112906293A (en) * | 2021-01-28 | 2021-06-04 | 北京航空航天大学 | Machine teaching method and system based on review mechanism |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101105853A (en) * | 2007-08-16 | 2008-01-16 | 上海交通大学 | Personalized teaching-guiding system based on non-zero jumping-off point in network teaching |
| CN103248693A (en) * | 2013-05-03 | 2013-08-14 | 东南大学 | Large-scale self-adaptive composite service optimization method based on multi-agent reinforced learning |
| CN104765842A (en) * | 2015-04-15 | 2015-07-08 | 中山大学 | Optimum study scheme pushing method and system |
| CN105700526A (en) * | 2016-01-13 | 2016-06-22 | 华北理工大学 | On-line sequence limit learning machine method possessing autonomous learning capability |
| US20170243507A1 (en) * | 2016-02-24 | 2017-08-24 | NEUROHM Sp. z o.o. Spolka komandytowa | Filters and related methods of use in measuring reaction times |
| CN108614865A (en) * | 2018-04-08 | 2018-10-02 | 暨南大学 | Method is recommended in individualized learning based on deeply study |
| CN109062919A (en) * | 2018-05-31 | 2018-12-21 | 腾讯科技（深圳）有限公司 | A kind of content recommendation method and device based on deeply study |
| CN109155049A (en) * | 2016-02-25 | 2019-01-04 | 日益得有限公司 | Method, equipment and the computer program of personal customization type education content are provided |
| CN109241424A (en) * | 2018-08-29 | 2019-01-18 | 陕西师范大学 | A kind of recommended method |
| CN109241291A (en) * | 2018-07-18 | 2019-01-18 | 华南师范大学 | Knowledge mapping optimal path inquiry system and method based on deeply study |
| CN109948054A (en) * | 2019-03-11 | 2019-06-28 | 北京航空航天大学 | A kind of adaptive learning path planning system based on intensified learning |
-
2019
- 2019-07-01 CN CN201910584394.2A patent/CN110288878B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101105853A (en) * | 2007-08-16 | 2008-01-16 | 上海交通大学 | Personalized teaching-guiding system based on non-zero jumping-off point in network teaching |
| CN103248693A (en) * | 2013-05-03 | 2013-08-14 | 东南大学 | Large-scale self-adaptive composite service optimization method based on multi-agent reinforced learning |
| CN104765842A (en) * | 2015-04-15 | 2015-07-08 | 中山大学 | Optimum study scheme pushing method and system |
| CN105700526A (en) * | 2016-01-13 | 2016-06-22 | 华北理工大学 | On-line sequence limit learning machine method possessing autonomous learning capability |
| US20170243507A1 (en) * | 2016-02-24 | 2017-08-24 | NEUROHM Sp. z o.o. Spolka komandytowa | Filters and related methods of use in measuring reaction times |
| CN109155049A (en) * | 2016-02-25 | 2019-01-04 | 日益得有限公司 | Method, equipment and the computer program of personal customization type education content are provided |
| CN108614865A (en) * | 2018-04-08 | 2018-10-02 | 暨南大学 | Method is recommended in individualized learning based on deeply study |
| CN109062919A (en) * | 2018-05-31 | 2018-12-21 | 腾讯科技（深圳）有限公司 | A kind of content recommendation method and device based on deeply study |
| CN109241291A (en) * | 2018-07-18 | 2019-01-18 | 华南师范大学 | Knowledge mapping optimal path inquiry system and method based on deeply study |
| CN109241424A (en) * | 2018-08-29 | 2019-01-18 | 陕西师范大学 | A kind of recommended method |
| CN109948054A (en) * | 2019-03-11 | 2019-06-28 | 北京航空航天大学 | A kind of adaptive learning path planning system based on intensified learning |
Non-Patent Citations (1)
| Title |
|---|
| 刘淇;陈恩红;朱天宇;黄振亚;吴润泽;苏喻;胡国平: "面向在线智慧学习的教育数据挖掘技术研究", 《模式识别与人工智能》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110928920A (en) * | 2019-11-19 | 2020-03-27 | 广东交通职业技术学院 | Knowledge recommendation method, system and storage medium based on improved position social contact |
| CN112053091A (en) * | 2020-09-28 | 2020-12-08 | 北京爱论答科技有限公司 | Data processing method and system based on learning operation |
| CN112180726A (en) * | 2020-09-29 | 2021-01-05 | 北京航空航天大学 | Spacecraft relative motion trajectory planning method based on meta-learning |
| CN112906293A (en) * | 2021-01-28 | 2021-06-04 | 北京航空航天大学 | Machine teaching method and system based on review mechanism |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110288878B (en) | 2021-10-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110288878A (en) | Adaptive learning method and device | |
| CN110941723A (en) | Method, system and storage medium for constructing knowledge graph | |
| CN111159419B (en) | Knowledge tracking data processing method, system and storage medium based on graph convolution | |
| CN106095812A (en) | Intelligent test paper generation method based on similarity measurement | |
| CN109242149A (en) | A kind of student performance early warning method and system excavated based on educational data | |
| CN109961098A (en) | A kind of training data selection method of machine learning | |
| CN116541538B (en) | Intelligent learning knowledge point mining method and system based on big data | |
| CN111461442B (en) | Knowledge tracking method and system based on federal learning | |
| CN112508334A (en) | Personalized paper combining method and system integrating cognitive characteristics and test question text information | |
| CN110046706A (en) | Model generating method, device and server | |
| CN113657723A (en) | Knowledge point mastering condition diagnosis method, test question recommendation method and device | |
| CN105205504B (en) | A kind of image attention regional quality evaluation index learning method based on data-driven | |
| Intisar et al. | Classification of online judge programmers based on rule extraction from self organizing feature map | |
| CN110968512A (en) | Software quality evaluation method, device, equipment and computer readable storage medium | |
| CN108509579A (en) | A kind of Intelligent exercising system and implementation method | |
| Chen et al. | Computer Adaptive Testing Using the Same-Decision Probability. | |
| CN114429212A (en) | Intelligent learning knowledge ability tracking method, electronic device and storage medium | |
| KR20100000017A (en) | An intelligent customized learning service method | |
| TW201919021A (en) | Method for determining and selecting knowledge node, and computer and storage medium for applying the same | |
| CN110222838A (en) | Deep neural network and its training method, device, electronic equipment and storage medium | |
| CN114490918A (en) | Dynamic knowledge graph building system and method based on dynamic learning model and application | |
| CN112541846B (en) | College course selection and repair mixed recommendation method and system based on attention mechanism | |
| CN114530073B (en) | Training method and device based on virtual reality | |
| US20220284315A1 (en) | Apparatus and system for training knowledge tracking model based on data augmentation and operation method thereof | |
| CN113673811B (en) | On-line learning performance evaluation method and device based on session |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |