Data-driven adaptive cognitive ability training method and device

Abstract

The invention discloses a data-driven self-adaptive cognitive competence training method and a data-driven self-adaptive cognitive competence training device. The method comprises the following steps: collecting initial cognitive competence data of a user; establishing a mapping relationship between the training tasks and the cognitive ability dimensions, and associating a plurality of predefined training tasks to the corresponding cognitive ability dimensions respectively; based on the user cognitive competence vector, selecting a training task matched with the current cognitive competence level of the user from a training task library through an adaptive recommendation algorithm; providing the training task matched with the current cognitive ability level of the user for the user to execute; and updating the user cognitive ability vector according to the behavior feedback data. According to the technical scheme, the training task is finely matched through the multi-dimensional cognitive vector, the difficulty is adjusted in real time, and the load is prevented from being too high or too low; and by utilizing behavior data closed-loop updating, a capability increasing curve can be accurately quantified, task sorting and rhythm are dynamically optimized, and a learning motivation is stimulated.

Description

Technical Field

[0001] The present invention relates to the technical field of cognitive training, and particularly to a data-driven adaptive cognitive ability training method and device. Background Art

[0002] With the in-depth research on neural plasticity and the popularization of the digital health industry, computerized cognitive training has become the main technical path to improve attention, working memory, and executive function. The academic and industrial communities have significantly improved training compliance and interest through gamification design, immersive interaction, and mobile deployment. To achieve individualized effects, multi-dimensional cognitive ability modeling and online learning algorithms have been introduced into the platform, and the system adjusts the task sequence, intensity, and feedback in real time according to the user's performance, thus forming a dynamic closed-loop adaptive training mechanism. At the same time, the combination of cloud computing and wearable sensors enables the platform to collect physiological data such as electroencephalogram and eye movement in real time, providing richer individual feature inputs for the algorithm.

[0003] However, existing platforms still have deficiencies. First, most products rely on the single increase or decrease of scores for grading, lacking a fine description of multi-dimensional ability differences, resulting in limited recommendation accuracy. Second, when the task difficulty mismatches with the real ability, it is easy to cause too high or too low cognitive load, reducing the training motivation. Third, the mapping between tasks and abilities is mostly set based on experience, lacking a weight model calibrated with behavioral data. Fourth, the utilization of fine-grained data such as click trajectories and reaction times is insufficient, and strategies cannot be adjusted immediately after detecting fatigue or distraction. Therefore, how to implement an adaptive cognitive training method with a multi-dimensional ability vector as the core, integrating real-time behavior feedback and dynamically adjusting training strategies has become an urgent issue in the industry. Summary of the Invention

[0004] The present invention provides a data-driven adaptive cognitive ability training method and device to achieve precise matching between tasks and abilities, as well as adaptive difficulty closed-loop control, improve training efficiency and motivation, reduce the cost of manual intervention, and promote continuous growth.

[0005] According to the first aspect of the present invention, there is provided a data-driven adaptive cognitive ability training method, which includes:

[0006] Collect the initial cognitive ability data of a user, define at least two independent cognitive ability dimensions, and construct a multi-dimensional user cognitive ability vector model;

[0007] Establish a mapping relationship between training tasks and the cognitive ability dimensions, associate a predefined plurality of training tasks with the corresponding cognitive ability dimensions respectively, and determine a corresponding cognitive ability requirement vector for each training task;

[0008] Based on the user's cognitive ability vector, select training tasks that match the user's current cognitive ability level from the training task library through an adaptive recommendation algorithm, where the reference factors of the adaptive recommendation algorithm include any one or more of the numerical distribution of the user's cognitive ability vector, the cognitive ability requirement vector of the training task, and the user's historical training data;

[0009] Provide the training tasks that match the user's current cognitive ability level for the user to execute, and collect the user's behavioral feedback data during the training in real time. The behavioral feedback data includes any one or more of the task completion progress, accuracy rate, reaction time, and operation process data;

[0010] Update the user's cognitive ability vector according to the behavioral feedback data to reflect the changes in each cognitive ability dimension of the user, and dynamically adjust the recommendation strategy and difficulty level of subsequent training tasks based on the updated user's cognitive ability vector and behavioral feedback data, so as to form an adaptive training closed loop based on user data feedback.

[0011] In one embodiment, the updating the user's cognitive ability vector according to the behavioral feedback data includes:

[0012] Analyze the performance data of the user in the training task, and compare the user's performance in each cognitive ability dimension involved in the training task with a preset benchmark level or difficulty expectation;

[0013] Calculate the update amount for each cognitive ability dimension according to the comparison result. The update amount is determined based on the deviation of the user's actual performance from the target difficulty;

[0014] Apply the update amount to the user's cognitive ability vector, adjust the value of the corresponding dimension, obtain the updated user's cognitive ability vector, and perform normalization or smoothing processing on the updated user's cognitive ability vector.

[0015] In one embodiment, the establishing the mapping relationship between the training task and the cognitive ability dimension includes:

[0016] Define the set of cognitive ability dimensions and clarify the cognitive ability categories represented by each dimension;

[0017] Based on expert experience and historical training data, determine the corresponding cognitive ability requirement vector for each pre-collected training task. The cognitive ability requirement vector includes the requirement weights of the training task in each cognitive ability dimension;

[0018] Associate and store each training task with its corresponding cognitive ability requirement vector, and construct a training task library to form a mapping relationship between the training task and the cognitive ability dimension.

[0019] In one embodiment, it further includes:

[0020] Obtain the user's current cognitive ability vector and its recent training history as input parameters for task recommendation;

[0021] According to the training task library and the mapping relationship, screen out several candidate training tasks from the training task library, where the candidate training tasks have a predetermined degree of complementarity or match with the user's cognitive ability vector in terms of the cognitive ability requirement vector;

[0022] Calculate the matching degree or expected benefit value of each candidate training task with the user's cognitive ability vector through an adaptive recommendation algorithm. The algorithm evaluates the potential effect of the task on improving the user's cognitive ability based on the differences between the numerical values of each dimension of the user's cognitive ability vector and the cognitive requirement vector of the candidate task;

[0023] Select at least one optimal task from the candidate training tasks according to the matching degree and recommend it to the user for execution, so that the recommended training task is adapted to the user's current cognitive ability level in terms of content and initial difficulty.

[0024] In one embodiment, it further includes:

[0025] For the training tasks completed by the user, evaluate the user's actual completion performance relative to the current difficulty level of the training task. The evaluation includes comparing indicators such as the user's accuracy rate and reaction time with the preset difficulty standard;

[0026] Based on the evaluation results, adjust the difficulty level of subsequent tasks according to preset rules;

[0027] Apply the adjusted difficulty level to the subsequently recommended training tasks, so that the difficulty level of the training task matches the user's real-time updated cognitive ability vector.

[0028] In one embodiment, it further includes:

[0029] Collect various behavior feedback data during the user's execution of the training task, including any one or more of the task completion time, number of errors, number of retries, click behavior, and attention concentration;

[0030] Analyze the behavior feedback data, evaluate the user's cognitive load and learning state changes, and identify the behavior characteristics of the user during the training process. The behavior characteristics include any one or more of fatigue degree, distraction degree, and specific error patterns;

[0031] Based on the analysis results of the behavior data, dynamically modify the type or order of the subsequently recommended training tasks, and adjust the rhythm of difficulty improvement.

[0032] According to a second aspect of the present invention, there is provided a data-driven adaptive cognitive ability training device, comprising:

[0033] An acquisition module, configured to acquire initial cognitive ability data of a user, define at least two independent cognitive ability dimensions, and construct a multi-dimensional user cognitive ability vector model;

[0034] A mapping module, configured to establish a mapping relationship between training tasks and the cognitive ability dimensions, associate a plurality of predefined training tasks with corresponding cognitive ability dimensions respectively, and determine a corresponding cognitive ability requirement vector for each training task;

[0035] An adaptive module, configured to select, based on the user cognitive ability vector, a training task that matches the user's current cognitive ability level from a training task library through an adaptive recommendation algorithm, wherein reference factors of the adaptive recommendation algorithm include any one or more of the numerical distribution of the user cognitive ability vector, the cognitive ability requirement vector of the training task, and the user's historical training data;

[0036] A providing module, configured to provide the training task that matches the user's current cognitive ability level for the user to execute, and collect real-time behavior feedback data of the user during the training process, where the behavior feedback data includes any one or more of task completion progress, correct rate, reaction time, and operation process data;

[0037] An updating module, configured to update the user cognitive ability vector according to the behavior feedback data to reflect changes in each cognitive ability dimension of the user, and dynamically adjust the recommendation strategy and difficulty level of subsequent training tasks based on the updated user cognitive ability vector and behavior feedback data, so as to form an adaptive training closed loop based on user data feedback.

[0038] In one embodiment, the acquisition module, the mapping module, the adaptive module, the providing module, and the updating module are controlled to implement any one of the above-mentioned data-driven adaptive cognitive ability training methods.

[0039] According to a third aspect of the present invention, there is provided an electronic device, which includes: a communication interface, a processor, and a memory;

[0040] Wherein, the memory is used to store program instructions, and when the program instructions are executed by the processor communicatively connected to the memory through the communication interface, any one of the above-mentioned data-driven adaptive cognitive ability training methods is implemented.

[0041] According to a fourth aspect of the present invention, there is provided a computer-readable storage medium having computer program instructions stored thereon, which when executed by a computer (e.g., a processor in the computer) implement any one of the above data-driven adaptive cognitive ability training methods.

[0042] In summary, the present invention provides a data-driven adaptive cognitive ability training method and apparatus. The method includes: collecting initial cognitive ability data of a user, defining at least two independent cognitive ability dimensions, and constructing a multi-dimensional user cognitive ability vector model; establishing a mapping relationship between training tasks and the cognitive ability dimensions, associating a plurality of predefined training tasks with corresponding cognitive ability dimensions respectively, and determining a corresponding cognitive ability requirement vector for each training task; based on the user cognitive ability vector, selecting a training task that matches the user's current cognitive ability level from a training task library through an adaptive recommendation algorithm, wherein reference factors of the adaptive recommendation algorithm include any one or more of the numerical distribution of the user cognitive ability vector, the cognitive ability requirement vector of the training task, and the user's historical training data; providing the training task that matches the user's current cognitive ability level for the user to execute, and collecting real-time behavior feedback data of the user during the training process, where the behavior feedback data includes any one or more of task completion progress, correct rate, reaction time, and operation process data; updating the user cognitive ability vector according to the behavior feedback data to reflect changes in each cognitive ability dimension of the user, and dynamically adjusting the recommendation strategy and difficulty level of subsequent training tasks based on the updated user cognitive ability vector and behavior feedback data, thereby forming an adaptive training closed loop based on user data feedback. The technical solution of this application finely matches training tasks through multi-dimensional cognitive vectors, adjusts the difficulty in real time, and avoids excessive or too low load; uses closed-loop update of behavior data to accurately quantify the ability growth curve, dynamically optimize task sorting and rhythm, and stimulate learning motivation; the combination of cloud computing and end-side sensing realizes a more consistent experience across devices, meets the needs of multiple scenarios such as precise rehabilitation and educational evaluation, and reduces the cost of manual intervention and maintenance.

[0043] Other features and advantages of the present invention will be described in the following specification, and in part will be obvious from the specification, or will be understood by implementing the present invention. The objectives and other advantages of the present invention can be realized and obtained by the structures specifically pointed out in the written specification and drawings.

[0044] The technical solution of the present invention will be further described in detail below through the drawings and embodiments. Description of the Drawings

[0045] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0046] Figure 1 Flowchart of a data-driven adaptive cognitive ability training method provided for an embodiment of the present invention;

[0047] Figure 2 Flowchart of another data-driven adaptive cognitive ability training method provided for an embodiment of the present invention;

[0048] Figure 3 Flowchart of another data-driven adaptive cognitive ability training method provided for an embodiment of the present invention;

[0049] Figure 4 Flowchart of another data-driven adaptive cognitive ability training method provided for an embodiment of the present invention;

[0050] Figure 5 Flowchart of another data-driven adaptive cognitive ability training method provided for an embodiment of the present invention;

[0051] Figure 6 Flowchart of another data-driven adaptive cognitive ability training method provided for an embodiment of the present invention;

[0052] Figure 7 Structural diagram of a data-driven adaptive cognitive ability training device provided for an embodiment of the present invention;

[0053] Figure 8 Structural diagram of an electronic device provided for an embodiment of the present invention. Specific embodiments

[0054] The following will describe in detail the features and exemplary embodiments of various aspects of the present application. To make the objectives, technical solutions, and advantages of the present application clearer, the following further describes the present application in detail with reference to the drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain the present application, rather than to limit the present application. For those skilled in the art, the present application can be implemented without some of these specific details. The following description of the embodiments is only to provide a better understanding of the present application by showing examples of the present application.

[0055] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprising", "including" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent in such process, method, article or device. Without further limitation, an element defined by the statement "comprising..." does not exclude the presence of additional identical elements in the process, method, article or device comprising the said element.

[0056] As Figure 1 shown, the present invention provides a data-driven adaptive cognitive ability training method, which includes:

[0057] In step S11, collect the initial cognitive ability data of the user, define at least two independent cognitive ability dimensions and construct a multi-dimensional user cognitive ability vector model;

[0058] In step S12, establish a mapping relationship between the training tasks and the cognitive ability dimensions, associate a plurality of predefined training tasks with the corresponding cognitive ability dimensions respectively, and determine a corresponding cognitive ability requirement vector for each training task;

[0059] In step S13, based on the user cognitive ability vector, select a training task that matches the user's current cognitive ability level from the training task library through an adaptive recommendation algorithm, wherein the reference factors of the adaptive recommendation algorithm include any one or more of the numerical distribution of the user cognitive ability vector, the cognitive ability requirement vector of the training task, and the user's historical training data;

[0060] In step S14, provide the training task that matches the user's current cognitive ability level for the user to execute, and collect the behavioral feedback data of the user during the training process in real time, where the behavioral feedback data includes any one or more of the task completion progress, correct rate, reaction time, and operation process data;

[0061] In step S15, update the user cognitive ability vector according to the behavioral feedback data to reflect the changes in each cognitive ability dimension of the user, and dynamically adjust the recommendation strategy and difficulty level of the subsequent training tasks based on the updated user cognitive ability vector and behavioral feedback data, so as to form an adaptive training closed loop based on user data feedback.

[0062] In one embodiment, with cognitive ability vector as the core control quantity, large-scale individual data evaluation and online adaptive recommendation are integrated to form a closed-loop structure of "evaluation-teaching-testing" that continuously cycles. The quantitative model accurately describes the user's initial ability state, and then introduces the training task demand vector to achieve multi-dimensional precise matching. The algorithm is driven by real-time behavioral data to self-correct, ensuring that subsequent task recommendations and difficulty adjustments always keep pace with the user's immediate ability changes, which can solve the problems of traditional cognitive training platforms in cold start, task mismatch, and effect evaluation.

[0063] The construction and initialization of the multidimensional ability vector is to collect the user's initial cognitive ability data and project it onto at least two independent cognitive ability dimensions, such as working memory, attention control, executive function, language processing or spatial reasoning. Arranging at least two independent cognitive ability dimensions in a preset order can form a high-dimensional vector model, and the value of each dimension reflects the user's current level in the corresponding cognitive subsystem. In order to avoid cold start bias, three types of data sources can be combined to standardize the pre-measurement table, short start-up task and previous historical records. The initial vector is smoothed and the confidence interval is estimated by algorithms such as least squares, expectation maximum or Kalman filtering.

[0064] The mapping of training tasks and capability dimensions is to provide a basis for the recommendation algorithm and to build a training task library in the background. Each training task is labeled by an expert group or assigned a set of weight values ​​through machine learning retrospective analysis. The weight represents the loading intensity of the task for each capability dimension. At the matrix level, the demand vector and the capability vector have the same dimensions, so the distance, angle or weighted difference in the vector space can be calculated. In order to further improve the objectivity and plasticity of the mapping, the task weights can be periodically re-evaluated through the continuously accumulated behavioral data. The mapping relationship is not a one-time static setting, but is gradually calibrated as the amount of platform data grows.

[0065] The decision-making mechanism of the adaptive recommendation algorithm is to input the user ability vector, task requirement vector and historical performance indicators into the recommendation engine at the same time during the algorithm decision-making stage. The recommendation engine contains both strategies that utilize exploration balance and prediction submodules based on gradient or Bayesian optimization to estimate the marginal benefits of each candidate task on overall ability improvement. The matching method gives priority to tasks that are highly consistent with the ability level to maintain training motivation; the gap-filling method selects tasks that are slightly higher than the current level in the user's weak dimension to create a moderate challenge. The system can switch freely between the above two strategies (matching and gap-filling). By adjusting the strategy parameters, the recommendation style can be flexibly set according to the training goals and user characteristics.

[0066] The real-time acquisition and analysis of behavioral feedback data is based on the system's fine-grained capture of multi-source behavioral data during task execution. Explicit metrics include task completion progress, accuracy rate, and reaction time; implicit metrics include click trajectories, pause distributions, retry counts, and even extensible physiological signals such as eye movements, heart rate, or galvanic skin response. The data first enters the stream processing pipeline to quickly complete outlier filtering and time window aggregation. The parsing engine uses corresponding models for different types of data. For example, the diffusion decision model is used to jointly explain the accuracy rate and reaction time, sequence clustering is used to identify continuous error patterns, and machine vision networks are used to determine distraction states triggered by line-of-sight deviation. The parsed results are mapped to an index system consistent with the ability dimension, and at the same time, macroscopic state estimates such as cognitive load, motivation level, and fatigue trend are output.

[0067] The dynamic update of the ability vector and difficulty adjustment is based on the update module comparing the behavioral analysis results with the current task difficulty benchmark to calculate the deviation in each dimension. The deviation can be converted based on the classical IRT formula or the non-linear relationship can be fitted through a neural network to adapt to complex interactive tasks. After obtaining the deviation, an update is immediately applied to the ability vector, and exponential moving average or Bayesian posterior smoothing is used to suppress single-point fluctuations. At the same time, the difficulty regulator recalibrates the initial difficulty level of the subsequent candidate tasks according to the challenge interval principle (i.e., making the center of the task difficulty distribution slightly higher than the user's ability level). If signs of overload are detected, the slope is reduced or consolidation tasks are inserted to ensure safety. At this time, the new ability vector and difficulty configuration jointly flow back to the recommendation engine to complete one iteration of the closed loop.

[0068] The technical solution in this embodiment finely matches the training tasks through multi-dimensional cognitive vectors, adjusts the difficulty in real time, and avoids excessive or too low load; uses the closed-loop update of behavioral data to accurately quantify the ability growth curve, dynamically optimize the task sorting and rhythm, and stimulate the learning motivation; the combination of cloud computing and end-side sensing realizes a more consistent experience across devices, meets the needs of multiple scenarios such as precise rehabilitation and education evaluation, and reduces the costs of manual intervention and maintenance. It can achieve high-precision ability estimation, fast-connected difficulty matching, and immediate and effective training rhythm control. Compared with the traditional fixed script or single-score promotion mode, it has high measurement fineness, which is reflected in that the multi-dimensional vector avoids information loss; it has a fast adaptation speed, which is reflected in that real-time calculation and stream processing enable the ability estimation to be refreshed synchronously as the task progresses; it takes both safety and motivation into account, which is reflected in reducing frustration and burnout through cognitive load monitoring and challenge interval adjustment. Further combined with the cooperation of Internet of Things devices, it can be implemented in scenarios such as home rehabilitation, elderly cognitive prevention, adolescent subject training, and workplace skill improvement.

[0069] In one embodiment, as Figure 2 shown, it further includes the following steps S21 - S23:

[0070] In step S21, analyze the performance data of the user in the training task, and compare the user's performance in each cognitive ability dimension involved in the training task with a preset benchmark level or difficulty expectation;

[0071] In step S22, calculate the update amount for each cognitive ability dimension according to the comparison result, and the update amount is determined based on the deviation of the user's actual performance relative to the target difficulty;

[0072] In step S23, apply the update amount to the user's cognitive ability vector, adjust the value of the corresponding dimension to obtain an updated user cognitive ability vector, and perform normalization or smoothing processing on the updated user cognitive ability vector.

[0073] In one embodiment, the multi-dimensional ability vector is the core state quantity of the entire platform. By comparing performance data with a benchmark, converting the deviation into an update amount, and correcting and refining the vector through vector correction and smoothing, the vector is continuously calibrated. Through the multi-dimensional item response theory, performance indicators such as accuracy rate and response time are mapped to a space with the same dimension as the ability vector, and then compared with the target difficulty given by experts or norms to generate a deviation for each dimension. According to the item loading coefficient and discrimination degree, the deviation is recalibrated into an accumulative gradient increment, and the Kalman filter is used to suppress noise. After adding the increment to the original vector, the peak-valley effect is eliminated through two-level processing of exponential moving average and global normalization, and when necessary, double-chain smoothing of fast and slow is achieved through the Kalman filter again. The closed-loop operation keeps the task difficulty always within the challenge range, promoting the targeted strengthening of weak items and the consolidation of strong items.

[0074] If the user completes the training task, the platform collects explicit data such as accuracy rate, response time, and completion progress, and at the same time captures implicit data such as click trajectories and pause distributions. The multi-dimensional item response theory provides a method to map these multi-source indicators into a performance vector with the same dimension, so that each numerical value can correspond one-to-one with the ability vector. Subsequently, the target difficulty is called as a ruler, and the difficulty can be dynamically updated through the population norm. After comparing the performance vector with the target difficulty for each dimension, the deviation is obtained. This deviation not only reveals whether the training session is too difficult or too easy, but also indicates whether a specific dimension needs to be strengthened or maintained.

[0075] Directly using the differential update is likely to cause distortion due to uneven item weights. Therefore, the platform introduces the item loading coefficient and discrimination degree into the formula, and first converts the deviation into a gradient increment with unified dimensions. In order to balance individual differences and long-term benefits in a dynamic scenario, a reinforcement learning strategy is adopted, regarding the immediate performance as a reward signal, and estimating the value of the increment and the policy change in real time. When external interference is large, the Kalman filter automatically adjusts the covariance in two stages of prediction and correction, weakening the impact of abnormal data on the gradient. For users with a long training sequence, the platform parallelly maintains the Bayesian knowledge tracing probability perspective to track the long-term mastery degree and supplement the time dimension information of the gradient.

[0076] The system accumulates the gradient increment dimension by dimension to the original ability vector to form a new vector. However, the accidental peak of a single task may lead to cliff-like fluctuations. Therefore, the platform first performs exponential moving average, assigning higher weights to the latest several updates, which not only tracks the real changes but also suppresses random noise. Subsequently, global normalization is performed to pull all dimensions back to a unified dimension scale, making it comparable among different users and different tasks. When a more delicate real-time trajectory is required in the scenario, the moving average output will enter the Kalman filter again to form a double-chain smoothing of fast response and slow trend. After the update is completed, the new ability vector is immediately fed back to the recommendation engine to drive the next round of task screening and difficulty setting, forming a second-level adaptive rhythm.

[0077] Multi-dimensional quantization enables the system to accurately distinguish fine-grained differences such as attention, memory, and executive control; the policy gradient combined with the Kalman filter completes the estimation and maintains smoothness within seconds, synchronizing the training rhythm with the user's state in real time; the cognitive load monitoring and the challenge interval theory jointly ensure that the training has sufficient pressure while avoiding frustration.

[0078] In one embodiment, as Figure 3 shown, it further includes the following steps S31 - S33:

[0079] In step S31, define the set of the cognitive ability dimensions and clarify the cognitive ability categories represented by each dimension;

[0080] In step S32, based on expert experience and historical training data, determine the corresponding cognitive ability requirement vector for each pre-collected training task. The cognitive ability requirement vector includes the requirement weights of this training task on each cognitive ability dimension;

[0081] In step S33, associate and store each training task with its corresponding cognitive ability requirement vector, and construct a training task library to form a mapping relationship between the training tasks and the cognitive ability dimensions.

[0082] In one embodiment, the cognitive training platform can achieve precise adaptation, establishing a computable mapping table between the training tasks and the ability dimensions. By delimiting the dimension set through cognitive science, generating requirement vectors for each task based on expert scoring and historical data, permanently binding the tasks and vectors and writing them into the library for the recommendation algorithm to quickly retrieve and dynamically calibrate. It has the measurement idea of multi-dimensional item response theory and also has the modeling paradigm of "entity - relationship - attribute" triples in knowledge graphs and machine learning.

[0083] Cognitive science usually breaks down core capabilities into subsystems such as working memory, inhibitory control, cognitive flexibility, and spatial reasoning. Independent evidence can be found in neuroimaging and behavioral experiments for each dimension, so they are independent of each other and measurable. After the dimension set is determined, a category definition needs to be given, that is, to state what kind of mental function each dimension represents and its observable indicators, which is equivalent to labeling the coordinate axes.

[0084] After determining the dimensions, each training task needs to be labeled with ability requirements, which is specifically manifested as a string of floating-point numbers of the same length as the dimensions, that is, the cognitive ability requirement vector. The platform usually adopts a two-step hybrid strategy. The first step is for neuroscience or educational psychology experts to give initial scores based on experimental literature and task design concepts to ensure theoretical rationality. The second step is to use historical training data to reverse calibrate the weights to ensure consistency with the actual improvement effect. Data-driven calibration can use a multidimensional item response model or a hybrid model of rough sets and fuzzy sets to compare the discrimination and improvement rates of the same task among subjects with different ability levels, so as to automatically adjust the weights of each dimension. After calibration, if a task has a weight of a for working memory and b for inhibitory control, and a > b, the recommendation algorithm can then judge that this task is more suitable for making up for memory shortfalls rather than control shortfalls, thereby improving the matching accuracy and reducing ineffective training.

[0085] Writing the training task-requirement vector pairs into the training task library is equivalent to preparing an ability routing table for the algorithm. Each record in the library contains at least fields such as task number, media form, requirement vector, current difficulty coefficient, and historical effect indicators, which is convenient for retrieval and sorting as needed. As the platform runs, the system will continuously collect the ability changes and behavioral loads of users after completing a certain task, and regularly fine-tune the requirement vector or difficulty coefficient according to actual data to achieve the self-evolution of the training task library. The closed loop of mapping-validation-re-mapping enables the training task library to correct itself with the latest evidence, avoiding the problem of the initial expert annotation aging over time, and also enabling the recommendation engine to output reliable decisions when facing new users and new scenarios.

[0086] The construction of the mapping relationship connects discrete training tasks and abstract cognitive dimensions into one. The dimension set determines what to measure, the requirement vector describes what to use for measurement, and the task library stores how to use it. It not only enables the platform to achieve fine evaluation and accurate recommendation, but also has the ability of dynamic evolution, and can automatically iterate with the upgrade of device sensing and the change of population characteristics, significantly improving the scientificity, adaptability and long-term maintainability of the cognitive training system.

[0087] In one embodiment, as Figure 4 shown, the following steps S41 - S44 are further included:

[0088] In step S41, obtain the user's current cognitive ability vector and its recent training history records as input parameters for task recommendation;

[0089] In step S42, according to the training task library and the mapping relationship, several candidate training tasks are screened out from the training task library, and the candidate training tasks are complementary or matched with the user cognitive ability vector to a predetermined degree in the cognitive ability requirement vector;

[0090] In step S43, the matching degree or expected benefit value between each candidate training task and the user cognitive ability vector is calculated through an adaptive recommendation algorithm, and the algorithm evaluates the potential effect of the task on improving the user's cognitive ability based on the differences between the numerical values of each dimension of the user cognitive ability vector and the cognitive requirement vector of the candidate task;

[0091] In step S44, at least one optimal task is selected from the candidate training tasks according to the matching degree and recommended to the user for execution, so that the recommended training task is adapted to the user's current cognitive ability level in terms of content and initial difficulty.

[0092] In one embodiment, the system first reads the user's multi-dimensional ability vector and recent training trajectories, and quickly locates matching or complementary candidates in the task library by means of a mapping table; then uses a reinforcement learning algorithm to estimate the potential benefits of each candidate task; finally, selects one to several optimal tasks according to the benefits, and sets an appropriate initial difficulty according to the current ability. The whole process not only absorbs the accuracy advantage of the multi-dimensional item response theory in ability quantification, but also utilizes the adaptive ability of reinforcement learning in real-time decision-making to ensure that the training is both personalized and interpretable.

[0093] Obtain the user's latest cognitive ability vector and the record of the most recent one or more trainings. The multi-dimensional ability vector is often generated by means of the multi-dimensional item response theory or its variants, which can provide measurement accuracy equivalent to that of traditional long scales within a short task sequence. Inputting the vector and historical scores into the recommendation engine can effectively alleviate the cold start problem, enabling the algorithm to still capture the user's preferences and weaknesses in the early stage without long-term observation. The pause distribution, error types and score trends in the historical records will also be analyzed as features to infer the learning motivation and cognitive load level, so as to avoid giving too high or too low initial difficulty.

[0094] Retrieve several tasks in the training task library whose cognitive requirement vector and ability vector are complementary or matched to a set threshold. The complementary strategy preferentially selects tasks with a higher load on the user's weak dimensions but still controllable overall difficulty for targeted shortcoming supplementation. The matching strategy, on the other hand, looks for tasks with a load on each dimension close to the user's ability to maintain a smooth experience and motivation. The task requirement vector is derived from expert annotation and historical efficacy calibration to ensure the objectivity and reliability of the load distribution. When the task library expands continuously with the operation of the platform, the mapping table will periodically re-evaluate the weights by means of incremental learning to ensure that the screening results are always close to the latest evidence.

[0095] The screened candidate tasks are sent to an adaptive recommendation algorithm for fine sorting. Research and industrial practice have shown that by adopting a multi-objective reinforcement learning framework, short-term performance and long-term gains can be tracked simultaneously in a real-time environment. The algorithm uses the difference vector, i.e., the per-dimensional difference between the user's ability and the task requirements, as a contextual feature, and then combines the interaction history, task diversity, and cognitive load prediction to calculate a comprehensive reward function. The reward not only measures the potential for ability improvement but also incorporates factors such as motivation maintenance, cognitive fatigue risk, and cross-dimensional balance. If the reward differences among different candidates are not significant, the strategy will moderately retain exploration to ensure that new tasks that may be more suitable for the user can be continuously discovered in the future.

[0096] Select at least one task with the highest reward and push it to the user, and set the starting difficulty level according to the current ability vector to ensure that the task falls within the optimal challenge range. During execution, metrics such as accuracy rate and reaction time will also be monitored. If signs of overload or burnout are detected, the difficulty can be immediately lowered or consolidation tasks can be inserted to reconcile the load. The task completion results and the updated ability vector are jointly written back to the database to form a closed-loop iteration of evaluation-recommendation-training-re-evaluation. Through this mechanism, the platform achieves second-level adaptability while ensuring interpretability, being able to accurately complement weaknesses and maintain the continuous motivation of users for long-term training.

[0097] In one embodiment, as Figure 5 shown, the following steps S51 - S53 are further included:

[0098] In step S51, define the set of cognitive ability dimensions and clarify the categories of cognitive abilities represented by each dimension;

[0099] In step S52, based on expert experience and historical training data, determine the corresponding cognitive ability requirement vector for each pre-collected training task. The cognitive ability requirement vector includes the requirement weights of the training task on each cognitive ability dimension;

[0100] In step S53, associate each training task with its corresponding cognitive ability requirement vector and store them to construct a training task library to form a mapping relationship between the training tasks and the cognitive ability dimensions.

[0101] In one embodiment, in the adaptive cognitive training system, difficulty adjustment ensures that the training is neither too difficult to cause frustration nor too easy to lead to burnout. Evaluate the user's actual performance on the current task, and according to the evaluation results, adjust the difficulty of subsequent tasks up or down according to preset rules, and write the new difficulty back to the recommendation engine so that the next batch of tasks is always within the optimal challenge range of the user's ability.

[0102] Perform multi-dimensional scoring on the task just completed by the user. The core indicators usually include accuracy rate and reaction time, which can capture both accuracy and processing speed simultaneously. Subsequently, these real-time data are compared with the task difficulty scale, which can be derived from a large-sample norm or the threshold can be set by experts according to the target training curve. If the accuracy rate is higher than the upper threshold and the reaction time is significantly faster than the norm, the system determines that the current difficulty is relatively low; if the accuracy rate is continuously lower than the lower threshold and the reaction time is significantly prolonged, it indicates that the difficulty is relatively high; being within the threshold range is considered a good match. The comparison mechanism and dynamic difficulty adjustment have been repeatedly empirically supported in gamified learning and executive function training, and can effectively improve motivation and transfer effects.

[0103] The evaluation results are sent to the decision-making module. The decision-making can use a simple threshold table for probability control to achieve the balance between exploration and exploitation. When the performance is better than the upper threshold for more than two rounds, the difficulty level is increased by one level; when the performance is worse than the lower threshold, the difficulty level is decreased by one level; when the performance fluctuates within the interval, the original difficulty level is maintained; at the same time, the maximum step size can be restricted to prevent excessive adjustment. Kalman filtering or exponential moving average is often used to smooth short-term fluctuations and avoid drastic jumps in difficulty caused by a single abnormal performance, so as to ensure the stability and predictability of the adjustment process. By parameterizing these rules, the system can also quickly switch strategy templates according to different population goals (rehabilitation, competition, academics).

[0104] The adjusted new difficulty is written back to the task record and input into the recommendation algorithm as a constraint condition, so that the next round of candidate tasks is sampled in the appropriate challenge interval. When the user enters the skill-challenge balance state, that is, the classic flow zone, both the learning efficiency and subjective pleasure will increase significantly. At the same time, the fine-tuning of the difficulty curve and the synchronous update of the ability vector form a two-way coupling. The amplitude of vector refresh determines the next task load, and the task completion data updates the vector in the reverse direction, realizing the cycle of evaluation-recommendation-training-re-evaluation. This closed-loop mechanism enables the platform to perceive and respond to changes in ability at the second level, continuously maintain the user in the optimal challenge interval, maximize the training benefit, and reduce the risk of cognitive fatigue through a smooth difficulty curve.

[0105] In one embodiment, as Figure 6 shown, it further includes the following steps S61 - S63:

[0106] In step S61, during the process of the user performing the training task, collect various behavior feedback data, including any one or more of the task completion time, the number of errors, the number of retries, click behavior, and attention concentration.

[0107] In step S62, analyze the behavior feedback data, evaluate the changes in the user's cognitive load and learning state, and identify the behavior characteristics of the user during the training process. The behavior characteristics include any one or more of fatigue degree, distraction degree, and specific error patterns.

[0108] In step S63, based on the analysis results of the behavior data, the type or order of the subsequent training task recommendations is dynamically modified, and the pace of increasing difficulty is adjusted.

[0109] In one embodiment, multi-channel behavioral data is collected during training, and these data are used to infer cognitive load, fatigue and attention status, and the task type, sequence and difficulty increase rhythm are adjusted in real time. Behavioral analysis, load detection and adaptive recommendation are tightly coupled to ensure training safety and maximize learning efficiency.

[0110] Cognitive load can be simultaneously characterized by explicit indicators (time, accuracy) and implicit indicators (click trajectory, eye concentration, number of retries). Extended completion time and increased error rate are often seen as direct signals of increased load or distracted attention. Continuous high load will cause cognitive fatigue, resulting in a decline in overall behavioral performance. At the same time, clickstream data can reveal distraction patterns such as free browsing and repeated operations, and is a fine-grained interaction clue commonly used in learning analysis research. If the platform is equipped with an eye or facial camera, the system can also calculate attention indicators such as gaze duration and scanning rate in the background to further improve state resolution. By aggregating these multimodal data in real time on the client or edge device, the system lays a multidimensional data foundation for subsequent load inference and rhythm control.

[0111] The collected data enters the analysis engine after stream processing. The platform performs differential operations on the time and accuracy with the preset standards and outputs the instantaneous load score. The high-load and low-load thresholds are set by norms or expert experience. Studies have shown that excessive load will significantly reduce learning transfer, while too low load will inhibit the motivation to challenge. The system identifies fatigue and distraction characteristics through sliding window statistics and sequence clustering. Lengthening of click intervals, repeated error patterns, and a sudden increase in the number of retries are early signs of fatigue. Eye movement deviation or wandering of the eyes can indicate a decrease in attention. Dedicated error analysis can also locate specific error patterns and attribute them to strategic defects or conceptual misunderstandings, providing accurate basis for subsequent remedial tasks. By mapping these behavioral characteristics to the dimensions of fatigue and distraction, the platform obtains a dynamic portrait of the learning state.

[0112] The state profile is sent to the decision layer and the recommendation layer for collaborative processing. If the system detects a continuous low load, it will give priority to tasks with higher cognitive demand vectors or more challenging question types in the task library to improve motivation and efficiency. If high load or fatigue growth is detected, consolidation tasks are inserted or the difficulty step is reduced to buffer pressure. The decision algorithm often uses a multi-armed bandit or its contextual extension to explore and utilize a balanced approach to maintain a personalized rhythm while avoiding local optimality. In order to avoid excessive fluctuations, the difficulty step will be processed by Kalman filtering or exponential smoothing to make the curve smooth and predictable. Feeding fatigue and distraction monitoring results directly to the difficulty regulator can significantly reduce premature dropout rates and improve long-term persistence.

[0113] Through the closed-loop of collection-inference-regulation, real-time load monitoring reduces the overload risk and protects user safety and experience. Multimodal behavior analysis enables the system to predict fatigue and distraction before a large number of errors occur, thus adjusting the rhythm in advance. The dynamic rearrangement of task types and sequences, combined with a gradual increase in difficulty, keeps users in a manageable challenge range, maximizing the flow and learning efficiency. The accumulated state-behavior-effectiveness data in turn trains the recommendation and prediction models, forming a positive feedback loop of continuous self-optimization. In summary, this behavior feedback-driven adaptive regulation mechanism combines cognitive science principles with machine learning techniques to provide a safe, effective, and interpretable dynamic rhythm management solution for the personalized training platform.

[0114] In one embodiment, Figure 7 is a block diagram of a data-driven adaptive cognitive ability training device shown according to an exemplary embodiment. As Figure 7 shown, the data-driven adaptive cognitive ability training device includes a collection module 71, a mapping module 72, an adaptive module 73, a providing module 74, and an updating module 75.

[0115] The collection module 71 is used to collect the initial cognitive ability data of the user, define at least two independent cognitive ability dimensions, and construct a multi-dimensional user cognitive ability vector model;

[0116] The mapping module 72 is used to establish a mapping relationship between the training tasks and the cognitive ability dimensions, associate a plurality of predefined training tasks with the corresponding cognitive ability dimensions respectively, and determine a corresponding cognitive ability requirement vector for each training task;

[0117] The adaptive module 73 is used to select, based on the user cognitive ability vector, a training task that matches the user's current cognitive ability level from the training task library through an adaptive recommendation algorithm, where the reference factors of the adaptive recommendation algorithm include any one or more of the numerical distribution of the user cognitive ability vector, the cognitive ability requirement vector of the training task, and the user's historical training data;

[0118] The providing module 74 is used to provide the training task that matches the user's current cognitive ability level for the user to execute, and collect the behavior feedback data of the user during the training process in real time, where the behavior feedback data includes any one or more of the task completion progress, correct rate, reaction time, and operation process data;

[0119] The update module 75 is used to update the user cognitive ability vector according to the behavior feedback data, so as to reflect the changes in each cognitive ability dimension of the user, and dynamically adjust the recommendation strategy and difficulty level of subsequent training tasks based on the updated user cognitive ability vector and behavior feedback data, thereby forming an adaptive training closed loop based on user data feedback.

[0120] The acquisition module 71, the mapping module 72, the adaptive module 73, the providing module 74, and the update module 75 included in the block diagram of the data-driven adaptive cognitive ability training device are controlled to execute the data-driven adaptive cognitive ability training method described in any of the above embodiments.

[0121] As Figure 8 shown, the present invention provides an electronic device 800, which includes: a communication interface, a processor 801, and a memory 802;

[0122] Among them, the memory 802 is used to store program instructions. When the program instructions are executed by the processor 801 communicatively connected to the memory 802 through the communication interface, the initial cognitive ability data of the user is collected, at least two independent cognitive ability dimensions are defined, and a multi-dimensional user cognitive ability vector model is constructed; a mapping relationship between the training tasks and the cognitive ability dimensions is established, a plurality of predefined training tasks are respectively associated with the corresponding cognitive ability dimensions, and a corresponding cognitive ability requirement vector is determined for each training task; based on the user cognitive ability vector, a training task matching the user's current cognitive ability level is selected from the training task library through an adaptive recommendation algorithm, where the reference factors of the adaptive recommendation algorithm include any one or more of the numerical distribution of the user cognitive ability vector, the cognitive ability requirement vector of the training task, and the user's historical training data; the training task matching the user's current cognitive ability level is provided for the user to execute, and the behavior feedback data of the user during the training process is collected in real time, and the behavior feedback data includes any one or more of the task completion progress, the correct rate, the reaction time, and the operation process data; the user cognitive ability vector is updated according to the behavior feedback data, so as to reflect the changes in each cognitive ability dimension of the user, and the recommendation strategy and difficulty level of subsequent training tasks are dynamically adjusted based on the updated user cognitive ability vector and behavior feedback data, thereby forming an adaptive training closed loop based on user data feedback.

[0123] The present invention provides a computer-readable storage medium, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the initial cognitive ability data of a user is collected, at least two independent cognitive ability dimensions are defined, and a multi-dimensional user cognitive ability vector model is constructed; a mapping relationship between training tasks and the cognitive ability dimensions is established, a plurality of predefined training tasks are respectively associated with the corresponding cognitive ability dimensions, and a corresponding cognitive ability requirement vector is determined for each training task; based on the user cognitive ability vector, a training task matching the current cognitive ability level of the user is selected from a training task library through an adaptive recommendation algorithm, wherein the reference factors of the adaptive recommendation algorithm include any one or more of the numerical distribution of the user cognitive ability vector, the cognitive ability requirement vector of the training task, and the historical training data of the user; the training task matching the current cognitive ability level of the user is provided for the user to execute, and the behavioral feedback data of the user during the training process is collected in real time, and the behavioral feedback data includes any one or more of task completion progress, correct rate, reaction time, and operation process data; the user cognitive ability vector is updated according to the behavioral feedback data to reflect the changes in each cognitive ability dimension of the user, and the recommendation strategy and difficulty level of subsequent training tasks are dynamically adjusted based on the updated user cognitive ability vector and the behavioral feedback data, so as to form an adaptive training closed loop based on user data feedback.

[0124] It should be understood that the specific features, operations, and details described above regarding the method of the present invention can also be similarly applied to the device and system of the present invention, or vice versa. In addition, each step of the method of the present invention described above can be executed by the corresponding components or units of the device or system of the present invention.

[0125] It should be understood that each module / unit of the device of the present invention can be implemented in whole or in part by software, hardware, firmware, or a combination thereof. Each module / unit can be embedded in the processor of a computer device in the form of hardware or firmware or independent of the processor, or can be stored in the memory of the computer device in the form of software for the processor to call to execute the operations of each module / unit. Each module / unit can be implemented as an independent component or module, or two or more modules / units can be implemented as a single component or module.

[0126] In one embodiment, a computer device is provided, which includes a memory and a processor. Computer instructions executable by the processor are stored on the memory. When the computer instructions are executed by the processor, the processor is instructed to execute the steps of the method according to the embodiments of the present invention. The computer device can be broadly a server, a terminal, or any other electronic device with necessary computing and / or processing capabilities. In one embodiment, the computer device may include a processor, a memory, a network interface, a communication interface, etc. connected through a system bus. The processor of the computer device can be used to provide necessary computing, processing, and / or control capabilities. The memory of the computer device may include a non-volatile storage medium and an internal memory. An operating system, a computer program, etc. may be stored in or on the non-volatile storage medium. The internal memory can provide an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface and the communication interface of the computer device can be used to connect and communicate with external devices through a network. When the computer program is executed by the processor, the steps of the method according to the present invention are executed.

[0127] The present invention can be implemented as a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method according to the embodiments of the present invention are caused to be executed. In one embodiment, the computer program is distributed on a plurality of network-coupled computer devices or processors, so that the computer program is stored, accessed, and executed in a distributed manner by one or more computer devices or processors. A single method step / operation, or two or more method steps / operations, can be executed by a single computer device or processor or by two or more computer devices or processors. One or more method steps / operations can be executed by one or more computer devices or processors, and one or more other method steps / operations can be executed by one or more other computer devices or processors. One or more computer devices or processors can execute a single method step / operation, or execute two or more method steps / operations.

[0128] Those of ordinary skill in the art can understand that the method steps of the present invention can be implemented by a computer program to instruct relevant hardware such as a computer device or a processor. The computer program can be stored in a non-transitory computer-readable storage medium, and when the computer program is executed, the steps of the present invention are caused to be executed. Depending on the circumstances, any reference herein to a memory, storage, database, or other medium may include non-volatile and / or volatile memory. Examples of non-volatile memory include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), flash memory, magnetic tape, floppy disk, magneto-optical data storage device, optical data storage device, hard disk, solid state disk, etc. Examples of volatile memory include random access memory (RAM), external cache memory, etc.

[0129] The technical features described above can be combined arbitrarily. Although not all possible combinations of these technical features are described, any combination of these technical features should be considered to be covered by this specification as long as such a combination does not result in a contradiction.

[0130] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features. And these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A data-driven adaptive cognitive ability training method, characterized in that, Including: Collecting the initial cognitive ability data of the user, defining at least two independent cognitive ability dimensions, and constructing a multi-dimensional user cognitive ability vector model; Establishing a mapping relationship between the training tasks and the cognitive ability dimensions, associating each of the predefined multiple training tasks with the corresponding cognitive ability dimension, and determining a corresponding cognitive ability requirement vector for each training task; Based on the user cognitive ability vector, selecting, through an adaptive recommendation algorithm, training tasks that match the user's current cognitive ability level from the training task library, wherein the reference factors of the adaptive recommendation algorithm include any one or more of the numerical distribution of the user cognitive ability vector, the cognitive ability requirement vector of the training task, and the user's historical training data; Providing the training tasks that match the user's current cognitive ability level for the user to execute, and collecting in real time the behavioral feedback data of the user during the training process, where the behavioral feedback data includes any one or more of the task completion progress, accuracy rate, reaction time, and operation process data; Updating the user cognitive ability vector according to the behavioral feedback data to reflect the changes in each cognitive ability dimension of the user, and dynamically adjusting the recommendation strategy and difficulty level of subsequent training tasks based on the updated user cognitive ability vector and behavioral feedback data, so as to form an adaptive training closed loop based on user data feedback.

2. The data-driven adaptive cognitive ability training method according to claim 1, wherein The updating the user cognitive ability vector according to the behavioral feedback data includes: Analyzing the performance data of the user in the training task, and comparing the user's performance on each cognitive ability dimension involved in the training task with a preset benchmark level or difficulty expectation; Calculating the update amount for each cognitive ability dimension according to the comparison result, where the update amount is determined based on the deviation of the user's actual performance from the target difficulty; Applying the update amount to the user's cognitive ability vector, adjusting the value of the corresponding dimension to obtain the updated user cognitive ability vector, and performing normalization or smoothing processing on the updated user cognitive ability vector.

3. The data-driven adaptive cognitive ability training method according to claim 1, wherein, The establishing the mapping relationship between the training tasks and the cognitive ability dimensions includes: Defining the set of the cognitive ability dimensions, and clarifying the cognitive ability categories represented by each dimension; Based on expert experience and historical training data, determining a corresponding cognitive ability requirement vector for each of the pre-collected training tasks, where the cognitive ability requirement vector includes the requirement weights of the training task on each cognitive ability dimension; Associating and storing each training task with its corresponding cognitive ability requirement vector, and constructing a training task library to form a mapping relationship between the training tasks and the cognitive ability dimensions.

4. The data-driven adaptive cognitive ability training method according to claim 3, wherein Also including: Obtaining the user's current cognitive ability vector and its recent training history record as input parameters for task recommendation; According to the training task library and the mapping relationship, screening out several candidate training tasks from the training task library, where there is a predetermined degree of complementarity or matching between the cognitive ability requirement vectors of the candidate training tasks and the user cognitive ability vector. Calculate the matching degree or expected benefit value between each candidate training task and the user's cognitive ability vector through an adaptive recommendation algorithm. The algorithm evaluates the potential effect of the task on enhancing the user's cognitive ability based on the differences between the numerical values of each dimension of the user's cognitive ability vector and the cognitive requirement vector of the candidate task. Select at least one optimal task from the candidate training tasks according to the matching degree and recommend it to the user for execution, so that the recommended training tasks are adapted to the user's current cognitive ability level in terms of content and initial difficulty.

5. The data-driven adaptive cognitive ability training method according to claim 4, wherein It also includes: For the training tasks completed by the user, evaluate the user's actual completion performance relative to the current difficulty level of the training task. The evaluation includes comparing indicators such as the user's accuracy rate and reaction time with the preset difficulty standard. Adjust the difficulty level of subsequent tasks according to the preset rules based on the evaluation results. Apply the adjusted difficulty level to the subsequently recommended training tasks, so that the difficulty level of the training tasks matches the user's real-time updated cognitive ability vector.

6. The data-driven adaptive cognitive ability training method according to claim 5, wherein It also includes: Collect various behavioral feedback data during the user's execution of the training task, including any one or more of the task completion time, number of errors, number of retries, click behavior, and attention concentration. Analyze the behavioral feedback data, evaluate the user's cognitive load and changes in the learning state, and identify the behavioral characteristics of the user during the training process. The behavioral characteristics include any one or more of fatigue, distraction, and specific error patterns. Based on the analysis results of the behavioral data, dynamically modify the type or order of the subsequently recommended training tasks, and adjust the rhythm of difficulty increase.

7. A device for data-driven adaptive cognitive ability training, characterized in that, It includes: A collection module for collecting the user's initial cognitive ability data, defining at least two independent cognitive ability dimensions, and constructing a multi-dimensional user cognitive ability vector model. A mapping module for establishing a mapping relationship between the training tasks and the cognitive ability dimensions, associating each of the predefined multiple training tasks with the corresponding cognitive ability dimension, and determining the corresponding cognitive ability requirement vector for each training task. An adaptive module for selecting training tasks that match the user's current cognitive ability level from the training task library through an adaptive recommendation algorithm based on the user's cognitive ability vector. Among them, the reference factors of the adaptive recommendation algorithm include any one or more of the numerical distribution of the user's cognitive ability vector, the cognitive ability requirement vector of the training task, and the user's historical training data. A providing module for providing the training tasks that match the user's current cognitive ability level for the user to execute, and collecting the behavioral feedback data of the user during the training process in real time. The behavioral feedback data includes any one or more of the task completion progress, correct rate, reaction time, and operation process data. An updating module for updating the user's cognitive ability vector according to the behavioral feedback data to reflect the changes in each cognitive ability dimension of the user, and dynamically adjusting the recommendation strategy and difficulty level of the subsequent training tasks based on the updated user's cognitive ability vector and behavioral feedback data, so as to form an adaptive training closed loop based on user data feedback.

8. The data-driven adaptive cognitive ability training device according to claim 7, wherein: The acquisition module, the mapping module, the adaptive module, the providing module, and the updating module are controlled to execute the data-driven adaptive cognitive ability training method according to any one of claims 1 to 6.

9. An electronic device, characterized in that, Comprising: A communication interface, a processor, and a memory; Wherein, the memory is used to store program instructions, and when the program instructions are executed by the processor communicatively connected to the memory through the communication interface, the electronic device is caused to implement the data-driven adaptive cognitive ability training method according to any one of claims 1 to 6.

10. A computer-readable storage medium having program instructions stored thereon, characterized in that, When the program instructions are executed by a computer, the computer is caused to implement the data-driven adaptive cognitive ability training method according to any one of claims 1 to 6.

Images