Single molecule conductivity identification method, device and equipment and computer readable storage medium

Abstract

The invention discloses a single-molecule conductance identification method, device and equipment and a computer readable storage medium, and relates to the technical field of single-molecule electrical measurement, and the method comprises the steps: obtaining source domain data, target domain data and a real conductance label of the source domain data; inputting the source domain data and the target domain data into a to-be-trained model to obtain respective conductance characteristics and predicted conductance labels of the source domain data and the target domain data; and optimizing the to-be-trained model based on each conductivity feature, each predicted conductivity tag and the real conductivity tag to obtain a new to-be-trained model, and returning to execute the step of inputting the source domain data and the target domain data into the to-be-trained model until the to-be-trained model meets a preset training termination condition to obtain a transfer learning model, and single-molecule conductivity state identification is carried out through the transfer learning model. According to the invention, the generalization ability of the model for identifying the conductivity state of the single molecule can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of single-molecule electrical measurement, and in particular to a single-molecule conductivity identification method, device, equipment and computer-readable storage medium. Background Art

[0002] With the continuous development of technologies such as scanning tunneling microscopy and mechanically controllable cleavage, researchers can now achieve electrical measurements at the single-molecule scale, revealing discrete conductivity peaks at the molecule-electrode interface. These phenomena correspond to different molecular conformations and connection states, which is of great significance for a deep understanding of the basic physical processes in molecular electronics.

[0003] At present, in the actual preparation process of single-molecule junctions, due to the influence of factors such as experimental environment and operation techniques, the single-molecule conductance traces obtained in different scenarios are quite different. Among them, some conductance traces are easy to classify and annotate to correspond to their conductance states. However, the model trained based on the annotated data set composed of limited conductance data is difficult to learn the global characteristics of the data and has the problem of low generalization ability.

[0004] Therefore, how to improve the generalization ability of the model used for single-molecule conductance state identification is a problem that needs to be solved urgently. Summary of the invention

[0005] The main purpose of this application is to provide a single-molecule conductance recognition method, device, equipment and computer-readable storage medium, aiming to improve the generalization ability of the model used for single-molecule conductance state recognition.

[0006] To achieve the above objectives, the present application provides a single molecule conductance recognition method, the single molecule conductance recognition method comprising: Acquire source domain data, target domain data, and a real conductance label of the source domain data, wherein the source domain data and the target domain data are single-molecule conductance data in different scenarios, and the real conductance label represents a real conductance state of a single molecule corresponding to the source domain data; Inputting the source domain data and the target domain data into the model to be trained to obtain the conductivity features and predicted conductivity labels of the source domain data and the target domain data respectively; The model to be trained is optimized based on each of the conductivity features, each of the predicted conductivity labels and the true conductivity label to obtain a new model to be trained, and the step of inputting the source domain data and the target domain data into the model to be trained is returned to be executed until the model to be trained meets a preset training termination condition, thereby obtaining a transfer learning model, so as to perform single-molecule conductance state recognition through the transfer learning model.

[0007] In one embodiment, the conductance feature includes a global conductance feature and a subdomain conductance feature, and the step of inputting the source domain data and the target domain data into the model to be trained to obtain the conductance features and predicted conductance labels of the source domain data and the target domain data respectively includes: Inputting the source domain data and the target domain data into the model to be trained, performing feature extraction on the source domain data and the target domain data through a feature extraction module of the model to be trained, and obtaining global conductivity features of the source domain data and the target domain data respectively; Performing dimensionality reduction processing on each of the global conductance features through the linear module of the model to be trained, so as to obtain subdomain conductance features of the source domain data and the target domain data respectively; The conductivity state of each of the subdomain conductivity features is classified by the classification module of the model to be trained to obtain the predicted conductivity labels of the source domain data and the target domain data.

[0008] In one embodiment, after the step of extracting features from the source domain data and the target domain data by the feature extraction module of the model to be trained to obtain the global conductivity features of the source domain data and the target domain data, the method further includes: The global adaptive module of the model to be trained is used to align the source domain data and the target domain data in terms of global spatial distribution based on the global conductivity features, so that the model to be trained can learn the global conductivity features of the target domain data.

[0009] In one embodiment, after the step of classifying the conductance state of each of the sub-domain conductance features through the classification module of the model to be trained to obtain the predicted conductance labels corresponding to the source domain data and the target domain data, the method further includes: The subdomain spatial distribution of the source domain data and the target domain data is aligned through the subdomain adaptation module of the model to be trained based on each of the subdomain conductance features, the predicted conductance label of the target domain data and the true conductance label, so that the model to be trained can learn the subdomain conductance features of the target domain data, wherein each subdomain space is composed of data corresponding to a type of conductance label.

[0010] In one embodiment, the step of optimizing the model to be trained based on each of the conductivity features, each of the predicted conductivity labels and the true conductivity label to obtain a new model to be trained includes: Determining a loss value of the model to be trained based on each of the conductance features, each of the predicted conductance labels, and the true conductance label; The model to be trained is optimized based on the loss value to obtain a new model to be trained.

[0011] In one embodiment, the step of determining the loss value of the model to be trained based on each of the conductance features, each of the predicted conductance labels and the true conductance label comprises: Determine a source domain classification loss value based on a first conductance label and the true conductance label, wherein the first conductance label is a predicted conductance label corresponding to the source domain data; Determining a domain adaptation loss value based on each of the conductance features, a second conductance label and the true conductance label, wherein the second conductance label is a predicted conductance label corresponding to the target domain data; The sum of the source domain classification loss value and the domain adaptation loss value is used as the loss value of the model to be trained.

[0012] In one embodiment, the domain adaptation loss value includes a global domain adaptation loss value and a subdomain adaptation loss value, and the step of determining the domain adaptation loss value based on each of the conductance features, the second conductance label and the real conductance label includes: Performing spatial distribution difference measurement on each of the global conductivity features to obtain the global adaptive loss value; The spatial distribution difference of each of the sub-domain conductivity features, the second conductivity label and the true conductivity label is measured to obtain the sub-domain adaptive loss value.

[0013] In addition, to achieve the above-mentioned purpose, the present application also provides a single molecule conductance recognition device, the single molecule conductance recognition device comprising: An acquisition module, used to acquire source domain data, target domain data and a real conductance label of the source domain data, wherein the source domain data and the target domain data are single-molecule conductance data in different scenarios, and the real conductance label represents the real conductance state of the single molecule corresponding to the source domain data; A prediction module, used to input the source domain data and the target domain data into a model to be trained to obtain the conductivity features and predicted conductivity labels of the source domain data and the target domain data respectively; A transfer learning module is used to optimize the model to be trained based on each of the conductivity features, each of the predicted conductivity labels and the true conductivity label to obtain a new model to be trained, and return to execute the step of inputting the source domain data and the target domain data into the model to be trained until the model to be trained meets a preset training termination condition, thereby obtaining a transfer learning model to perform single-molecule conductivity state recognition through the transfer learning model.

[0014] In addition, to achieve the above-mentioned purpose, the present application also provides a storage medium, which is a computer-readable storage medium, and the computer-readable storage medium stores a program for implementing the single-molecule conductance recognition method. The program for implementing the single-molecule conductance recognition method is executed by a processor to implement the steps of the single-molecule conductance recognition method as described above.

[0015] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, including a computer program, which implements the steps of the single-molecule conductance recognition method as described above when executed by a processor.

[0016] The present application provides a single-molecule conductance identification method. The present application first obtains source domain data, target domain data and a real conductance label of the source domain data, wherein the source domain data and the target domain data are single-molecule conductance data in different scenarios, and the real conductance label represents the real conductance state of the single molecule corresponding to the source domain data; the source domain data and the target domain data are input into a model to be trained to obtain conductance features and predicted conductance labels corresponding to the source domain data and the target domain data respectively; then the model to be trained is optimized based on each conductance feature, each predicted conductance label and the real conductance label to obtain a new model to be trained, and the step of inputting the source domain data and the target domain data into the model to be trained is returned to execute until the model to be trained meets a preset training termination condition, and a transfer learning model is obtained, so as to perform single-molecule conductance state identification through the transfer learning model.

[0017] In summary, the present application inputs the source domain data marked with real conductivity labels and the target domain data not marked with conductivity labels into the model to be trained, analyzes the source domain data and the target domain data through the model to be trained to obtain the conductivity characteristics of the data, and predicts the conductivity labels of the single molecules corresponding to the source domain data and the target domain data (i.e., predicts the conductivity labels), and then optimizes the model based on the conductivity characteristics and the prediction results until the model meets the preset training termination conditions, and obtains a transfer learning model for identifying the conductivity state of the single molecule. In this way, compared with the traditional method of only training the model based on labeled conductivity data, which leads to weak generalization ability of the model, the present application uses transfer learning to enable the model to use the knowledge learned from the source domain data to identify and classify the target domain data, thereby improving the generalization ability of the model for single-molecule conductivity state recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0019] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0020] Figure 1 This is a schematic diagram of the process of the first embodiment of the single-molecule conductance recognition method of the present application; Figure 2 A schematic diagram showing a comparison of conductivity traces involved in an embodiment of the single-molecule conductivity recognition method of the present application; Figure 3 A schematic diagram of an attention mechanism involved in an embodiment of the single-molecule conductance recognition method of the present application; Figure 4 A schematic diagram of target domain data classification and recognition results involved in an embodiment of the single-molecule conductance recognition method of the present application; Figure 5 A schematic diagram of a model loss calculation process involved in an embodiment of the single-molecule conductance recognition method of the present application; Figure 6 A schematic diagram of global adaptation involved in an embodiment of the single-molecule conductance recognition method of the present application; Figure 7 A schematic diagram of subdomain adaptation involved in an embodiment of the single-molecule conductance recognition method of the present application; Figure 8 A schematic diagram of a single molecule conductance recognition process involved in an embodiment of the single molecule conductance recognition method of the present application; Fig. 9 This is a schematic diagram of the module structure of the single-molecule conductance recognition device of this application; Fig.10 Schematic diagram of the device structure of the hardware operating environment involved in the single-molecule conductance recognition method in the embodiment of the present application.

[0021] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0022] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0023] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0024] The main solution of the present application is: obtaining source domain data, target domain data and real conductance labels of the source domain data, wherein the source domain data and the target domain data are single-molecule conductance data in different scenarios, and the real conductance label represents the real conductance state of the single molecule corresponding to the source domain data; inputting the source domain data and the target domain data into the model to be trained to obtain the conductance characteristics and predicted conductance labels of the source domain data and the target domain data respectively; optimizing the model to be trained based on each of the conductance characteristics, each of the predicted conductance labels and the real conductance label to obtain a new model to be trained, and returning to execute the step of inputting the source domain data and the target domain data into the model to be trained until the model to be trained meets the preset training termination condition, and obtaining a transfer learning model, so as to perform single-molecule conductance state recognition through the transfer learning model.

[0025] At present, in the actual preparation process of single-molecule junctions, due to the influence of factors such as experimental environment and operation techniques, the single-molecule conductance traces obtained in different scenarios are quite different. Among them, some conductance traces are easy to classify and annotate to correspond to their conductance states. However, the model trained based on the annotated data set composed of limited conductance data is difficult to learn the global characteristics of the data and has the problem of low generalization ability.

[0026] Therefore, how to improve the generalization ability of the model used for single-molecule conductance state identification is a problem that needs to be solved urgently.

[0027] This application inputs source domain data labeled with real conductivity labels and target domain data not labeled with conductivity labels into the model to be trained, analyzes the source domain data and target domain data through the model to be trained to obtain the conductivity characteristics of the data, and predicts the conductivity labels of single molecules corresponding to the source domain data and target domain data (i.e., predicting the conductivity labels), and then optimizes the model based on the conductivity characteristics and the prediction results until the model meets the preset training termination conditions, thereby obtaining a transfer learning model for identifying the conductivity state of single molecules. In this way, compared with the traditional method of only training models based on labeled conductivity data, which leads to weak model generalization ability, this application uses transfer learning to enable the model to use the knowledge learned from the source domain data to identify and classify the target domain data, thereby improving the generalization ability of the model for single-molecule conductivity state recognition.

[0028] It should be noted that the execution subject of the method in each embodiment of the single-molecule conductance recognition method of the present application can be a single-molecule conductance recognition system, or a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or a single-molecule conductance recognition device capable of realizing the above functions, etc., which is not specifically limited in this embodiment. The following takes the single-molecule conductance recognition system as an example to illustrate this embodiment and the following embodiments.

[0029] Based on this, the present application proposes a single molecule conductance recognition method of the first embodiment, please refer to Figure 1 , the single molecule conductivity recognition method comprises steps S10 to S30: Step S10, obtaining source domain data, target domain data and a real conductance label of the source domain data, wherein the source domain data and the target domain data are single-molecule conductance data in different scenarios, and the real conductance label represents a real conductance state of a single molecule corresponding to the source domain data; It should be noted that domain adaptation technology is a branch of transfer learning. In domain adaptation, there are two basic concepts, namely domain and task. Among them, the domain is the subject of transfer learning, including the source domain (SourceDomain) and the target domain (Target Domain). The source domain is a domain with knowledge and data labels, and the target domain is a domain that needs to be given data labels. The task is the goal of transfer learning. By performing the task to train a classifier, the knowledge learned in the source domain is transferred to the target domain to achieve the correct labeling of the sample data in the target domain. The source domain data refers to the knowledge in the source domain, and the real conductivity label of the source domain data refers to the data label of the source domain. The target domain data refers to the sample data in the target domain. The source domain data and the target domain data are single-molecule conductivity data in different scenarios. The real conductivity label represents the real conductivity state of the single molecule corresponding to the source domain data. Among them, the real conductivity label includes conductivity and non-conductivity.

[0030] In this embodiment, data preparation is first performed, that is, obtaining multiple source domain data and the true conductivity label of each source domain data , and obtain multiple target domain data . The number of source domain data is recorded as , the number of target domain data is recorded as .

[0031] Exemplarily, there are multiple different scenarios where the source domain data and the target domain data are located, as shown in the following table: Domain Adaptive Scenario Table

[0032] In a first feasible implementation, the source domain data and the target domain data correspond to different experimental environments, that is, the source domain data is the conductance data of molecule X in experimental environment A, and the target domain data is the conductance data of molecule X in experimental environment B; in a second feasible implementation, the source domain data and the target domain data correspond to different calculation methods, that is, the source domain data is the conductance data obtained by simulating and calculating molecule X in a computational simulation software, and the target domain data is the conductance data obtained by actually measuring the conductance of molecule X in a laboratory; in a third feasible implementation, the source domain data and the target domain data correspond to different molecules, that is, the source domain data is the conductance data of molecule X in an experimental environment, and the target domain data is the conductance data of molecule Y in an experimental environment.

[0033] For example, the conductivity traces (i.e., conductivity data) of a molecule are collected in experimental environment A and experimental environment B (there are differences in vibration noise and electrical equipment between the experiments, etc.). The conductivity data collected in experimental environment A is called source domain data. Source domain data is usually conductivity data with clear and obvious conductivity traces, low noise, and easy to classify and mark. The conductivity data collected in experimental environment B is called target domain data, and the target domain data is usually conductivity data with large fluctuations in the conductivity traces. Figure 2 The following is a schematic diagram of the comparison of conductivity traces, where: Figure 2 (a) shows the conductance-free step trace of a single molecule in the source domain (labeled as non-conductance). Figure 2 (b) shows the conductance step trace of a single molecule in the source domain (labeled as conductance), and Figure 2 (c) shows the conductivity trace of a single molecule in the target domain. In the data preparation stage, the source domain data is represented as , the total sample size of source domain data is ,Label It is divided into non-conductance steps (i.e., no conductance) and conductance steps (i.e., conductance). The target domain data is represented as , the total sample size of the target domain data is Since the number of points of each conductivity trace collected during the experiment is different, the fixed-length sampling method is used to process the conductivity traces so that the conductivity data has a unified dimension, which is convenient for subsequent model processing. The specifications of the processed source domain data are , the target domain data specification is .

[0034] Step S20, inputting the source domain data and the target domain data into a model to be trained to obtain conductivity features and predicted conductivity labels of the source domain data and the target domain data respectively; It should be noted that the embodiment of the present application does not limit the specific type of the model to be trained. In this embodiment, the model to be trained can be a convolutional neural network model or a recurrent network model.

[0035] The source domain data and the target domain data are input into the model to be trained, and the model to be trained performs feature extraction on the source domain data and the target domain data to obtain the conductivity features of the data, and performs classification prediction on the source domain data and the target domain data to obtain predicted conductivity labels. The predicted conductivity labels include conductivity and non-conductivity.

[0036] It should be understood that when there are multiple source domain data inputs and multiple target domain data inputs, the model to be trained will obtain the conductivity features and predicted conductivity labels of each input data.

[0037] Step S30, optimizing the model to be trained based on each of the conductivity features, each of the predicted conductivity labels and the true conductivity label to obtain a new model to be trained, and returning to execute the step of inputting the source domain data and the target domain data into the model to be trained until the model to be trained meets the preset training termination condition, and obtaining a transfer learning model, so as to perform single-molecule conductance state recognition through the transfer learning model.

[0038] It should be noted that the model training termination condition is pre-set as the number of training steps reaching a preset number of steps, or the model loss function converges.

[0039] Based on the conductance features, predicted conductance labels and true conductance labels output by the model to be trained, the model parameters of the model to be trained are optimized to obtain a new model to be trained, and the source domain data and the target domain data are re-input into the new model to be trained until the optimized model to be trained meets the preset training termination conditions. The model to be trained at this time is used as a transfer learning model for single-molecule conductance state recognition.

[0040] It should be noted that in traditional machine learning methods: the threshold method relies on prior baseline information and is sensitive to Gaussian white noise; the clustering algorithm has limited dimensionality reduction capabilities for high-dimensional features and is difficult to distinguish similar conductivity states (such as the difference between upright and tilted conformations); the accuracy of the SVM (Support Vector Machine) classifier is highly dependent on manual features (such as peak-to-valley ratio and autocorrelation coefficient), and features need to be redesigned when the molecular system changes. In conventional deep learning models: due to the scarcity of labeled samples, training on small sample data sets is prone to overfitting, and the model cannot adapt to the differences in conductivity trace data of the same molecule due to changes in the experimental environment (current noise, conductivity length and height, etc.). For example, in the stability test of single-molecule devices, the model excessively memorizes the noise pattern in the training data, resulting in a decrease in the recognition accuracy of new data.

[0041] The embodiments of the present application use domain adaptation technology to solve the cross-domain generalization problem of classification models in single-molecule conductance data analysis for different scenarios in areas such as differences in experimental conditions, differences in molecular systems, and differences in noise patterns. Through feature alignment, sample weighting, and model optimization, the robustness of the model in different experimental conditions and molecular systems is significantly improved. Compared with traditional methods (traditional machine learning models, early deep learning models), the use of domain adaptation technology can improve the generalization ability of classification models across experimental conditions while reducing the degree of dependence on labeled data, significantly improving the robustness, efficiency, and interpretability of single-molecule conductance data recognition, helping researchers to more easily obtain valuable electrical feature information from the data, and providing guidance for further theoretical research and experimental design.

[0042] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above-mentioned embodiment 1 can be referred to the above introduction, and will not be repeated later. On this basis, the conductivity feature includes the global conductivity feature and the sub-domain conductivity feature, and the step S20 may include: Step S201, inputting the source domain data and the target domain data into a model to be trained, performing feature extraction on the source domain data and the target domain data through a feature extraction module of the model to be trained, and obtaining global conductivity features of the source domain data and the target domain data respectively; It should be noted that the conductivity feature includes the global conductivity feature. The model to be trained includes a feature extraction module for extracting the conductivity features in the source domain data and the target domain data. The conductivity features include but are not limited to the conductivity height, conductivity length and conductivity distribution.

[0043] After the source domain data and the target domain data are input into the model to be trained, the feature extraction module of the model to be trained is used to extract features of the source domain data and the target domain data to obtain the low-dimensional features (i.e., global conductivity features) of the source domain data and the target domain data respectively.

[0044] Step S202, performing dimensionality reduction processing on each of the global conductance features through the linear module of the model to be trained to obtain subdomain conductance features of the source domain data and the target domain data; It should be noted that the conductivity feature also includes a subdomain conductivity feature, and the dimension of the subdomain conductivity feature is lower than that of the global conductivity feature. Exemplarily, the dimension of the global conductivity feature is 64 dimensions, and the dimension of the subdomain conductivity feature is 32 dimensions. The model to be trained also includes a linear module for performing feature dimensionality reduction processing on the global conductivity feature to obtain a low-dimensional feature (i.e., the subdomain conductivity feature) of a lower dimension.

[0045] Step S203, classifying the conductivity state of each of the sub-domain conductivity features through the classification module of the model to be trained, and obtaining the predicted conductivity labels of the source domain data and the target domain data respectively.

[0046] It should be noted that the model to be trained also includes a classification module for outputting sample labels (ie, predicted conductance labels).

[0047] In one feasible implementation, after the data preparation step, a convolutional neural network or a recurrent network is used to build the model, and an attention mechanism module is introduced into the model to improve the model's weight distribution of the conductivity feature, that is, a one-dimensional convolutional neural network with an attention mechanism is used to build a feature extraction module, denoted as Module_1; a linear module is connected after the feature extraction module to further reduce the dimension of the feature, denoted as Module_2; a classification module is connected after the linear module to output the sample label, denoted as Module_3. The total model Model = Module_1+Module_2+Module_3. The source domain data and the target domain data are respectively obtained through Module_1. Feature vectors and , that is, the global conductivity characteristics; and Then pass through Module_2 to get the eigenvector and , i.e., the subdomain conductivity characteristics; and Get the predicted label through Module_3 and , i.e., predicting the conductance label.

[0048] For example, Figure 3 The diagram shows an attention mechanism. The source domain data and the target domain data are input into a model with an attention mechanism. The model extracts the conductivity feature in the data as the input feature F, and then performs maximum pooling and average pooling on the input feature F. The processed data is input into the shared perception layer, and finally a weight is assigned to each channel, that is, a weight is assigned to each conductivity feature. In this way, the embodiment of the present application introduces an attention mechanism into the domain adaptation technology, so that the model is easy to focus on the key features of the data, thereby enhancing the interpretability of the model through visualization operations.

[0049] In this embodiment, the step S30 may include: Step S301, determining a loss value of the model to be trained based on each of the conductance features, each of the predicted conductance labels and the true conductance label; Step S302: Optimize the model to be trained based on the loss value to obtain a new model to be trained.

[0050] Based on the conductance features of each source domain data and each target domain data output by the model to be trained, the predicted conductance label and the true conductance label of each source domain data, the loss value of the model to be trained in the current round of training is calculated, and the parameters in the model to be trained are optimized based on the loss value to obtain the optimized model to be trained.

[0051] Exemplarily, after calculating the loss value of this round of model training, the optimizer can be used to reversely optimize the weight parameters of each module in the model to be trained until the preset step threshold is reached or the loss value is less than the preset threshold (ie, the loss function converges), and the model training is confirmed to be complete. Among them, the optimizer can be SGD (Stochastic Gradient Descent) or adaptive gradient methods. Then, the trained classification model (ie, transfer learning model) is set as a gradient-free optimization model, and the target domain data is input into the classification model again to obtain the predicted conductivity label to complete the classification recognition task. In addition, the embodiment of the present application can also use Python language to build a model for user terminal recognition analysis, or C++ language to build a model in an embedded device to complete a specific automated recognition task according to actual needs.

[0052] In this embodiment, the step S301 may include: Step A10, determining a source domain classification loss value based on a first conductance label and the true conductance label, wherein the first conductance label is a predicted conductance label corresponding to the source domain data; It should be noted that the predicted conductance label corresponding to the source domain data is called the first conductance label for distinction.

[0053] In one possible implementation, the cross entropy loss method is used to calculate the true conductivity label of the source domain data. and predicted conductivity labels The classification loss between (i.e., the source domain classification loss value) can be expressed as:

[0054] in, is the source domain classification loss value, represents the cross entropy loss calculation. It can be understood that when the number of source domain data is When , the source domain classification loss is calculated based on the true conductance label and predicted conductance label of each source domain data.

[0055] Step A20, determining a domain adaptation loss value based on each of the conductance features, the second conductance label and the true conductance label, wherein the second conductance label is a predicted conductance label corresponding to the target domain data; It should be noted that the predicted conductance label corresponding to the target domain data is called the second conductance label for distinction.

[0056] Based on each conductance feature output by the model to be trained, the second conductance label and the true conductance label, a domain adaptation loss value is calculated.

[0057] In this embodiment, the domain adaptation loss value includes a global domain adaptation loss value and a subdomain adaptation loss value, and the step A20 may include: Step A201, measuring the spatial distribution difference of each of the global conductivity features to obtain the global adaptive loss; It should be noted that the conductivity feature includes the global conductivity feature, and the domain adaptive loss value includes the global adaptive loss value.

[0058] In one feasible implementation, the MMD (Maximum Mean Discrepancy) method based on RKHS (Reproducing Kernel Hilbert Space) is used to calculate the feature vectors of the source domain data and the target domain data. and The global adaptive loss between , the calculation formula is:

[0059] Step A202, performing spatial distribution difference measurement on each of the sub-domain conductivity features, the second conductivity label and the true conductivity label to obtain the sub-domain adaptive loss value.

[0060] It should be noted that the conductance feature includes the subdomain conductance feature, and the domain adaptive loss value includes the subdomain adaptive loss value.

[0061] In one possible implementation, the feature vector of the source domain data is calculated using the MMD method based on the reproducing kernel Hilbert space. and real conductance labels , the feature vector of the target domain data and the subdomain adaptation loss between the predicted conductance labels , the calculation formula is:

[0062] Wherein, c represents the type of the conductivity label. In this embodiment, the value of c is 2.

[0063] Step A30: taking the sum of the source domain classification loss value and the domain adaptation loss value as the loss value of the model to be trained.

[0064] It should be noted that the loss value of the model to be trained includes the source domain classification loss ( ) and domain adaptation loss ( ), where there are two types of domain adaptation losses: global domain adaptation loss ( ) and subdomain adaptation loss ( ). The maximum mean difference method (MMD) is used to measure the spatial distribution difference. The following is the specific calculation method:

[0065]

[0066] The total loss function of the model to be trained is , According to the total loss value, the model is trained in reverse until the model meets the preset training termination condition to obtain the classification model. Then the classification model is tested based on the target domain data, that is, the classification and recognition task of the target domain data is completed.

[0067] For example, Figure 4 The following is a schematic diagram of the target domain data classification and recognition results. Figure 4 (a) is the target domain data to be classified and identified, including multiple conductance traces. The target domain data is input into the classification model, and the classification model can classify each conductance trace into a trace without conductance step and a trace with conductance step, that is, Figure 4 (b) is the conductance-free step trace of a single molecule in the target domain. Figure 4 (c) shows the conductance step trace of a single molecule in the target domain.

[0068] In one possible implementation, if Figure 5 The figure shows the model loss calculation process. First, the source domain data and the target domain data are input into the convolutional neural network with attention mechanism (i.e., feature extraction module) to obtain low-dimensional features. , that is, the global conductivity feature; then the low-dimensional feature is transformed into Perform dimensionality reduction to obtain low-dimensional features , that is, the subdomain conductivity feature; then the model classification layer (i.e., classification module) is used to classify the low-dimensional features The predicted labels of the source domain data and the target domain data (i.e., predicted conductance labels) are obtained. The global loss (i.e., global adaptive loss value) is calculated based on the global conductance features, the subdomain loss (i.e., subdomain adaptive loss value) is calculated based on the conductance features of each subdomain, the true label of the source domain data, and the predicted label of the target domain data, and the classification loss (i.e., source domain classification loss value) is calculated based on the predicted label of the source domain data and the true label of the source domain data. The above three losses are added together to obtain the total loss of the model.

[0069] In this way, the embodiment of the present application measures the difference between the source domain and the target domain by designing two types of loss functions (source domain classification loss and domain adaptation loss), and minimizes this difference during the model training process, so that the source domain data and the target domain data are closer in the feature space, thereby improving the effect of model adaptation.

[0070] In this embodiment, after step S201, the single molecule conductance recognition method of the present application further includes: Step B10, performing global spatial distribution alignment of the source domain data and the target domain data based on the global conductivity features through the global adaptive module of the model to be trained, so that the model to be trained can learn the global conductivity features of the target domain data.

[0071] It should be noted that the model to be trained also includes a global adaptation module, which is used to perform global adaptation operations on source domain data and target domain data.

[0072] After the feature extraction module of the model to be trained is used to extract features from the source domain data and the target domain data to obtain the global conductivity features of each data, the global adaptive module is used to align the overall numerical space distribution of the source domain data and the target domain data based on the global conductivity features, so that the model to be trained can learn the global conductivity features of the target domain.

[0073] For example, Figure 6 The figure shows a schematic diagram of global adaptation, which fuses the source domain data and the target domain data into a numerical distribution space, that is, realizes the numerical space distribution alignment. It can be understood that the blue figure represents the source domain data and the yellow figure represents the target domain data.

[0074] In this embodiment, after step S203, the single molecule conductance recognition method of the present application further includes: Step B20, through the subdomain adaptation module of the model to be trained, based on each of the subdomain conductance features, the predicted conductance label of the target domain data and the true conductance label, the subdomain space distribution of the source domain data and the target domain data is aligned, so that the model to be trained can learn the subdomain conductance features of the target domain data, wherein each subdomain space is composed of data corresponding to a type of conductance label.

[0075] It should be noted that the model to be trained also includes a subdomain adaptation module, which is used to perform subdomain adaptation operations on the source domain data and the target domain data.

[0076] After the model to be trained outputs the conductivity features of each subdomain and the predicted conductivity labels of the target domain data, the subdomain adaptive module aligns the subdomain numerical space distribution of the space after global adaptive alignment one by one according to different conductivity label categories.

[0077] For example, Figure 7 The figure shows a schematic diagram of subdomain adaptation. It can be understood that the triangles and squares represent different types of conductivity labels, respectively. The source domain data and the target domain data are divided into two subdomains according to different categories, and the numerical space distribution of the subdomains is aligned. That is, each subdomain space includes data corresponding to the same type of conductivity label.

[0078] In one possible implementation, if Figure 8 The figure shows a schematic diagram of the single-molecule conductance recognition process. First, data preparation is performed to obtain source domain conductance data (i.e., source domain data) and target domain conductance data (i.e., target domain data); the source domain conductance data and the target domain conductance data are input into the model to obtain the model output; the source domain classification loss and the domain adaptation loss are calculated based on the model output; the model parameters are reversely optimized based on the sum of the two types of losses until the optimized model meets the training termination conditions to obtain a classification model; the target domain conductance samples (i.e., target domain data) are predicted by the classification model to obtain the conductance labels of the target domain conductance samples.

[0079] In this way, the embodiment of the present application uses domain adaptation technology to transfer the knowledge learned in the source domain to the target domain to achieve correct labeling of target domain samples, effectively bridging the differences between conductivity data in different scenarios, so that the model trained in one domain can be better applied to other domains, improving the generalization ability and applicability of the model. At the same time, since it is costly to obtain a large amount of high-quality annotated data in single-molecule conductivity research, the embodiment of the present application realizes the learning of unknown label data in the target domain with the help of a small amount of annotated data in the source domain. Through transfer learning, the dependence on the annotated data in the target domain is reduced, thereby achieving more efficient use of existing data resources for accurate identification and classification, and improving the model's ability to handle complex distributed data.

[0080] The present application also provides a single molecule conductivity recognition device, please refer to Fig. 9 , the single-molecule conductance recognition device comprises: An acquisition module 10 is used to acquire source domain data, target domain data and a real conductivity label of the source domain data, wherein the source domain data and the target domain data are single-molecule conductivity data in different scenarios, and the real conductivity label represents the real conductivity state of the single molecule corresponding to the source domain data; A prediction module 20, configured to input the source domain data and the target domain data into a model to be trained, and obtain the conductivity features and predicted conductivity labels of the source domain data and the target domain data respectively; The transfer learning module 30 is used to optimize the model to be trained based on each of the conductivity features, each of the predicted conductivity labels and the true conductivity labels to obtain a new model to be trained, and return to execute the step of inputting the source domain data and the target domain data into the model to be trained until the model to be trained meets the preset training termination condition, thereby obtaining a transfer learning model to perform single-molecule conductivity state recognition through the transfer learning model.

[0081] Optionally, the conductivity feature includes a global conductivity feature and a subdomain conductivity feature, and the prediction module 20 is further used for: Inputting the source domain data and the target domain data into the model to be trained, performing feature extraction on the source domain data and the target domain data through a feature extraction module of the model to be trained, and obtaining global conductivity features of the source domain data and the target domain data respectively; Performing dimensionality reduction processing on each of the global conductance features through the linear module of the model to be trained, so as to obtain subdomain conductance features of the source domain data and the target domain data respectively; The conductivity state of each of the subdomain conductivity features is classified by the classification module of the model to be trained to obtain the predicted conductivity labels of the source domain data and the target domain data.

[0082] Optionally, the single-molecule conductance recognition device further comprises a global adaptive module, and the global adaptive module is used to: The global adaptive module of the model to be trained is used to align the source domain data and the target domain data in terms of global spatial distribution based on the global conductivity features, so that the model to be trained can learn the global conductivity features of the target domain data.

[0083] Optionally, the single-molecule conductance recognition device further comprises a subdomain adaptation module, and the subdomain adaptation module is used to: The subdomain spatial distribution of the source domain data and the target domain data is aligned through the subdomain adaptation module of the model to be trained based on each of the subdomain conductance features, the predicted conductance label of the target domain data and the true conductance label, so that the model to be trained can learn the subdomain conductance features of the target domain data, wherein each subdomain space is composed of data corresponding to a type of conductance label.

[0084] Optionally, the transfer learning module 30 is further used for: Determining a loss value of the model to be trained based on each of the conductance features, each of the predicted conductance labels, and the true conductance label; The model to be trained is optimized based on the loss value to obtain a new model to be trained.

[0085] Optionally, the transfer learning module 30 is further used for: Determine a source domain classification loss value based on a first conductance label and the true conductance label, wherein the first conductance label is a predicted conductance label corresponding to the source domain data; Determining a domain adaptation loss value based on each of the conductance features, a second conductance label and the true conductance label, wherein the second conductance label is a predicted conductance label corresponding to the target domain data; The sum of the source domain classification loss value and the domain adaptation loss value is used as the loss value of the model to be trained.

[0086] Optionally, the domain adaptive loss value includes a global domain adaptive loss value and a subdomain adaptive loss value, and the transfer learning module 30 is further used for: Performing spatial distribution difference measurement on each of the global conductivity features to obtain the global adaptive loss value; The spatial distribution difference of each of the sub-domain conductivity features, the second conductivity label and the true conductivity label is measured to obtain the sub-domain adaptive loss value.

[0087] The single-molecule conductance recognition device provided in the embodiment of the present application adopts the single-molecule conductance recognition method in the above embodiment, which can solve the technical problem of how to improve the generalization ability of the model for single-molecule conductance state recognition. Compared with the prior art, the beneficial effects of the single-molecule conductance recognition device provided in the embodiment of the present application are the same as the beneficial effects of the single-molecule conductance recognition method provided in the above embodiment, and other technical features in the single-molecule conductance recognition device are the same as the features disclosed in the above embodiment method, which will not be repeated here.

[0088] The present application provides a single-molecule conductance recognition device, which includes: at least one processor; and a memory that is communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the single-molecule conductance recognition method in the above-mentioned embodiment 1.

[0089] Reference below Fig.10, which shows a schematic diagram of the structure of a single-molecule conductance recognition device suitable for implementing the embodiments of the present application. The single-molecule conductance recognition device in the embodiments of the present application may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), etc., and fixed terminals such as digital TVs, desktop computers, etc. Fig.10 The single-molecule conductivity recognition device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0090] like Fig.10 As shown, the single-molecule conductance recognition device may include a processing device 1001 (such as a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to the program stored in the read-only memory 1002 or the program loaded from the storage device 1003 to the random access memory 1004. In the random access memory 1004, various programs and data required for the operation of the single-molecule conductance recognition device are also stored. The processing device 1001, the read-only memory 1002, and the random access memory 1004 are connected to each other through a bus 1005. The input / output interface 1006 is also connected to the bus. Generally, the following systems can be connected to the input / output interface 1006: an input device 1007 including, for example, a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD: Liquid Crystal Display), a speaker, a vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 can allow the single-molecule conductivity recognition device to communicate with other devices wirelessly or by wire to exchange data. Although the single-molecule conductivity recognition device with various systems is shown in the figure, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems can be implemented or have alternatively.

[0091] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a read-only memory 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0092] The single-molecule conductance recognition device provided by the present application adopts the single-molecule conductance recognition method in the above embodiment, which can solve the technical problem of how to improve the generalization ability of the model used for single-molecule conductance state recognition. Compared with the prior art, the beneficial effects of the single-molecule conductance recognition device provided by the present application are the same as the beneficial effects of the single-molecule conductance recognition method provided by the above embodiment, and other technical features in the single-molecule conductance recognition device are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0093] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0094] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0095] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer program) stored thereon, and the computer-readable program instructions are used to execute the single-molecule conductance recognition method in the above-mentioned embodiment.

[0096] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM: Random Access Memory), a read-only memory (ROM: Read Only Memory), an erasable programmable read-only memory (EPROM: Erasable Programmable Read Only Memory or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM: CD-Read Only Memory), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency: Radio Frequency), etc., or any suitable combination of the above.

[0097] The computer-readable storage medium may be included in the single-molecule conductance recognition device; or may exist independently without being assembled into the single-molecule conductance recognition device.

[0098] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the single-molecule conductance recognition device, the single-molecule conductance recognition device: obtains source domain data, target domain data and a real conductance label of the source domain data, wherein the source domain data and the target domain data are single-molecule conductance data in different scenarios, and the real conductance label represents the real conductance state of the single molecule corresponding to the source domain data; inputs the source domain data and the target domain data into the model to be trained to obtain the conductance characteristics and predicted conductance labels of the source domain data and the target domain data respectively; optimizes the model to be trained based on each of the conductance characteristics, each of the predicted conductance labels and the real conductance label to obtain a new model to be trained, and returns to execute the step of inputting the source domain data and the target domain data into the model to be trained until the model to be trained meets the preset training termination condition, thereby obtaining a transfer learning model, so as to perform single-molecule conductance state recognition through the transfer learning model.

[0099] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0100] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0101] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0102] The readable storage medium provided in the present application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned single-molecule conductance recognition method, and can solve the technical problem of how to improve the generalization ability of the model for single-molecule conductance state recognition. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in the present application are the same as the beneficial effects of the single-molecule conductance recognition method provided in the above-mentioned embodiment, and will not be repeated here.

[0103] An embodiment of the present application provides a computer program product, including a computer program, which implements the steps of the single-molecule conductance recognition method as described above when executed by a processor.

[0104] The computer program product provided in this application can improve the generalization ability of the model for single-molecule conductance state recognition. Compared with the prior art, the beneficial effects of the computer program product provided in the embodiment of this application are the same as the beneficial effects of the single-molecule conductance recognition method provided in the above embodiment, which will not be repeated here.

[0105] The above are only preferred embodiments of the present application, and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent processing scope of the present application.

Claims

1. A single molecule conductivity recognition method, characterized in that: The single molecule conductance recognition method comprises: Acquire source domain data, target domain data, and a real conductance label of the source domain data, wherein the source domain data and the target domain data are single-molecule conductance data in different scenarios, and the real conductance label represents a real conductance state of a single molecule corresponding to the source domain data; Inputting the source domain data and the target domain data into the model to be trained to obtain the conductivity features and predicted conductivity labels of the source domain data and the target domain data respectively; The model to be trained is optimized based on each of the conductivity features, each of the predicted conductivity labels and the true conductivity label to obtain a new model to be trained, and the step of inputting the source domain data and the target domain data into the model to be trained is returned to be executed until the model to be trained meets a preset training termination condition, thereby obtaining a transfer learning model, so as to perform single-molecule conductance state recognition through the transfer learning model.

2. The single molecule conductivity recognition method according to claim 1, characterized in that: The conductance feature includes a global conductance feature and a subdomain conductance feature. The step of inputting the source domain data and the target domain data into the model to be trained to obtain the conductance features and predicted conductance labels of the source domain data and the target domain data respectively includes: Inputting the source domain data and the target domain data into the model to be trained, performing feature extraction on the source domain data and the target domain data through a feature extraction module of the model to be trained, and obtaining global conductivity features of the source domain data and the target domain data respectively; Performing dimensionality reduction processing on each of the global conductance features through the linear module of the model to be trained, so as to obtain subdomain conductance features of the source domain data and the target domain data respectively; The conductivity state of each of the sub-domain conductivity features is classified by the classification module of the model to be trained to obtain the predicted conductivity labels of the source domain data and the target domain data.

3. The single molecule conductivity recognition method according to claim 2, characterized in that: After the step of extracting features from the source domain data and the target domain data by the feature extraction module of the model to be trained to obtain global conductivity features of the source domain data and the target domain data, the method further includes: The global adaptive module of the model to be trained is used to align the source domain data and the target domain data in terms of global spatial distribution based on the global conductivity features, so that the model to be trained can learn the global conductivity features of the target domain data.

4. The single molecule conductivity recognition method according to claim 2, characterized in that: After the step of classifying the conductance state of each of the subdomain conductance features through the classification module of the model to be trained to obtain the predicted conductance labels corresponding to the source domain data and the target domain data, the method further includes: The subdomain spatial distribution of the source domain data and the target domain data is aligned through the subdomain adaptation module of the model to be trained based on each of the subdomain conductance features, the predicted conductance label of the target domain data and the true conductance label, so that the model to be trained can learn the subdomain conductance features of the target domain data, wherein each subdomain space is composed of data corresponding to a type of conductance label.

5. The single molecule conductivity recognition method according to claim 2, characterized in that: The step of optimizing the model to be trained based on each of the conductivity features, each of the predicted conductivity labels and the true conductivity label to obtain a new model to be trained comprises: Determining a loss value of the model to be trained based on each of the conductance features, each of the predicted conductance labels, and the true conductance label; The model to be trained is optimized based on the loss value to obtain a new model to be trained.

6. The single molecule conductivity recognition method according to claim 5, characterized in that: The step of determining the loss value of the model to be trained based on each of the conductance features, each of the predicted conductance labels and the true conductance label comprises: Determine a source domain classification loss value based on a first conductance label and the true conductance label, wherein the first conductance label is a predicted conductance label corresponding to the source domain data; Determining a domain adaptation loss value based on each of the conductance features, a second conductance label and the true conductance label, wherein the second conductance label is a predicted conductance label corresponding to the target domain data; The sum of the source domain classification loss value and the domain adaptation loss value is used as the loss value of the model to be trained.

7. The single molecule conductivity recognition method according to claim 6, characterized in that: The domain adaptation loss value includes a global domain adaptation loss value and a subdomain adaptation loss value. The step of determining the domain adaptation loss value based on each of the conductance features, the second conductance label and the real conductance label includes: Performing spatial distribution difference measurement on each of the global conductivity features to obtain the global adaptive loss value; The spatial distribution difference of each of the sub-domain conductivity features, the second conductivity label and the true conductivity label is measured to obtain the sub-domain adaptive loss value.

8. A single molecule conductivity recognition device, characterized in that: The single-molecule conductance recognition device comprises: An acquisition module, used to acquire source domain data, target domain data and a real conductance label of the source domain data, wherein the source domain data and the target domain data are single-molecule conductance data in different scenarios, and the real conductance label represents the real conductance state of the single molecule corresponding to the source domain data; A prediction module, used to input the source domain data and the target domain data into a model to be trained to obtain the conductivity features and predicted conductivity labels of the source domain data and the target domain data respectively; A transfer learning module is used to optimize the model to be trained based on each of the conductivity features, each of the predicted conductivity labels and the true conductivity label to obtain a new model to be trained, and return to execute the step of inputting the source domain data and the target domain data into the model to be trained until the model to be trained meets a preset training termination condition, thereby obtaining a transfer learning model to perform single-molecule conductivity state recognition through the transfer learning model.

9. A single molecule conductivity recognition device, characterized in that: The single-molecule conductance recognition device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the single-molecule conductance recognition method according to any one of claims 1 to 7.

10. A storage medium, characterized in that: The storage medium is a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the single-molecule conductance recognition method according to any one of claims 1 to 7 are implemented.

Images