U - Ollama Platform Knowledge Graph

Today, I will introduce and demonstrate how to create a CI model to recommend a tourist spot using the QCI and AI learning platform. So, let's get started. First, sign in with your account and the password. Next, click on "create the CM model button," then enter a name for your CI model, for example, travel recommendation system. And then, next provide a description of your CI model. And next, on this web page we, need to set the number of model variable. We have three input variables and one output variable. And, on this webpage, we need to select the variable type. We have fuzzy variable and TSK variable for the Mamdani inference model, please select 'fuzzy variable' as your variable type. For TSK inference model, select fuzzy variable as your input variable type. In this video, we use Mamdani inference model as an example, so we choose fuzzy varable for our all of the variables. So, the first variable name is Express the second name is evaluation, and the third one variable name is the distance and the output varable name is recommendation level. And, we save them. This page enables you to set values for each attribute of every variable. For all variables, we have selected 'Maximum' as the accumulation method and COG as defuzzifier method. The first varable name is expense and its scale is Taiwanese dollars, with a boundary value ranging from 0 to 1. It indicates the amount you want to spend while traveling to this tourist spot. For the second variable is Evaluation, its scale is point and its boundary is from to 100. And, it reflects the level of evaluation rating given by the people who have visited this place. The third variable name is Distance. We use kilometers as its scale and because we are in Taiwan so we set its boundary is from 0 to 500. And, the output name is recommendation level, its scale is point, and its boundary is from 0 to 100. And, click on Update button after setting. Next, we need to set linguistic term for each variable. First, we set the linguistic term for variable Expense. [Music] After that, check the expense membership function shape. And then, next, we need to set the linguistic term for the second variable Evaluation. [Music] After that, check evaluation membership function shape. Next, we need to set the linguistic term for variable [Music] Distance. After that, check Distance membership function shape. So, next, we need to set the linguistic term for variable Recommendation Level. [Music] After that, check Recommendation Level membership function shape. After you finish parameter settings for all the variables' lingustic terms, you can click on Lingustic Term Functions button in this page and you will see all functions for your variables expense, evalation, distance, and recommendation level. And, you also can choose download knowledge model functions together and this is 'download all the membership functions' together and you also can choose download them one by one. This is expense. And, download evaluation, download distance, and download recommendation level. Congratulations! You already finished creating your knowledge model of your real-world applications so see you next time.

Hello, my name is Mei-Hui Wang, and I come from the National University of Tainan, Taiwan. Today, I will introduce and demonstrate how to create a CI Inference model for Travel Recommendation System using the QCI&AI-FML Learning Platform. So, let's get started. First, we need to download the template file, and then open the template file to create inference rules for this model. So, after finishing the construction for 27 rules, we need to save them This is the constructed inference rules. In this model, we set three input variables:  Expense, Evaluation and Distance, each with three linguistic terms. These variables and terms cause the inference model to have 27 (fuzzy) rules, one for each combination of the input variables. For example, check the first rule: If Expense is Low, Evaluation is Bad and Distance is Near, then the RecommendationLevel is "Not Recommended". However, the linguistic term for the "THEN" part may vary among different people, as individuals have unique perspectives and interpretations, influenced by their own judgment and intuition. After finishing construction, you can Click on the Load Inference Model button to load the inference model. Then, you can see the 5-layer inference model structure. The first layer is the input layer, including three variables, Expense, Evaluation, and Distance. The last layer is the output layer, variable RecommendationLevel. You have the option to download images of model structure viewed from bottom to top and from left to right. It also supports to download the inference model. Additionally, you also can click on "Clear Inference Model" to reset the model. So, we have now finished constructing the inference model using the QCI&AI-FML Learning Platform. Congratulations!! See you next time. Bye-Bye.

Hello, my name is Mei-Hui Wang, and I come from the National University of Tainan, Taiwan. Today, I will introduce and demonstrate how to conduct CI Inference for Travel Recommendation System using the QCI&AI-FML Learning Platform. So, let's get started. You can add data one by one by clicking on the "Add Data" button. For example, Expense is 714 TWD, Evaluation is 19, Distance is 422 km. And click on blue arrow button, the inferred result for RecommendationLevel is 17.619. That is, when Expense is High, Evaluation is Bad, Distance is Far, the QCI&AI-FML Learning Platform does not recommend visiting this place. You can load the inference data by clicking on this button. But first, we need to download template. Then, once you have finished creating the inference data, please save it. After constructing, the data we load the inference data. Here is the constructed inference data after loading Let's take the first data point as an example: Expense is 255, Evaluation is 72, and Distance is 197 km, What value and recommendation level will the QCI&AI-FML Learning Platform provide? Click on this blue arrow button to infer only this data. You also can click on "Infer" button to infer data in a batch. Consider the first data: When Expense is Low, Evaluation is Good, Distance is Normal. The QCI&AI-FML Learning Platform recommends you to visit this place very much. Additionally, you have the option to download inference result with semantics or download inference result without semantics. You also can click on "Clear Inference Data" to reset the data. So, we have now finished conducting CI inference using the QCI&AI-FML Learning Platform. Congratulations!! See you next time. Bye-Bye.

Hello, my name is Mei-Hui Wang, and I come from the National University of Tainan, Taiwan. Today, I will introduce and demonstrate how to learn expert knowledge for a Travel Recommendation System using the QCI&AI-FML Learning Platform. So, let's get started. You can load the expert data by clicking on this button. But, first we need to download the template file. Once you have completed creating the data, please save them. After creating expert data, we can load expert data. Here is the expert data. Click on this "Infer" button to process 500 data in a batch. The QCI&AI-FML Learning Platform employs semantic matching accuracy to verify its consistency with expert interpretations and uses MSE(Mean Square Error) and RMSE(Root Mean Square Error) to provide insights into the model's prediction error. After learning from the expert's knowlege, the total semantic matching accuracy is 0.882, MSE is 294.943 RMSE is 17.174. Let's take the first data as an example: When Expense is 255 , Evaluation is 72, Distance is 197, the Learning Results is 82.381. However, the experts justify a value of 70, resulting in a square error value of 153.288. But, the semantics match each other. This suggests that while the values are different, they might be conveying a similar underlying assessment or conclusion. Additionally, you have the option to download the learning results. You also can click on "Clear" to reset data. So, we now have finished CI model learning using the QCI&AI-FML Learning Platform Congratulations!! See you next time. Bye-Bye.

Hello, my name is Mei-Hui Wang, and I come from the National University of Tanan, Taiwan. Today, I will introduce and demonstrate how to train the model for a Travel Recommendation System using QCI&AI-FML Learning Platform. So, let's get started. The QCI&AI-FML Learning Platform provides the Particle Swarm Optimization(PSO) method to train the knowledge model of the CI model. PSO is the type of evolutionary computation technique, inspired by the collective behavior of animal swarms or flocks in nature. The population in PSO is composed of a number of particles. Each particle's movement is influenced by the best position it has found and the best position discovered by the swarm. This process helps in finding the optimal solution. Let's get started to train the model. You can load data by clicking on "Load Training Data" button. However, if your training data are not yet ready, first download template file to create the data. Once you have finished creating the data, please save it. After creating the dataset, please load the training data. Then, set the number of particles, (For example: 10) the number of iterations, (For example: 1000) And then, click on "Train Model" button to start training. During training, a message box will display updates, including iteration numbers, MSE, RMSE, and semantic matching accuracy. Additionally, a red circle at the top-right of this message box will flash. Upon completion of the training, a message box will appear indicating that the training has ended. Then, please click on "Download Results" button to download the training results. Additionary, please click on "Save Trained Model" button to save the trained model into the archived CI models by inputting your CI model name and CI model description. For example here (TrainedModel_TRS_1000Iter_10Ps) and click on "Save" button. At the top of this page, semantic matching accuracy, MSE and RMSE values are displayed after training. The curves of MSE and RMSE indicate a tendency to decrease. You can download all curves together or save each curve individually. For example: Save accuracy curve, MSE curve, RMSE curve. This page displays the knowledge model before training on the left-hand side and the knowledge model after training on the right-hand side. You can see that the membership function shape has changed to bring the results closer to the expert's knowledge. Additionally, you can download the pre-training knowledge model functions all together or save them individually. For example: Expense before training, Evaluation (before training), Distance (before training), RecommendationLevel (before training). The same options are available for the post-training model functions. For example: download all together and download individually Expense, Evaluation, Distance, RecommendationLevel. Finally, if you wish to access the trained model, you can find it in the archived CI models. Simply click on "Download" to download the trained model, or click on "Edit" button, then click "Next" button, then "Linguistic Term Functions" button to view the trained knowledge model. Congratulations on successfully completing CI machine learning using the QCI&AI-FML Learning Platform!! See you next time. Bye-Bye.

Hello, my name is Mei-Hui Wang, and I come from the National University of Tainan, Taiwan. Today, I will introduce and demonstrate how to save CI model for a Travel Recommendation System using QCI&AI-FML Learning Platform. So, let's get started. The QCI&AI-FML Learning Platform enables you to save your CI model as a new model. You also can click on "Download CI Model" to your local computer. Additionally, you also can load your CI model to validate the model to meet the IEEE 1855 standard. Additionary, you can directly upload the model to the QCI&AI-FML Learning Platform by clicking on "Create New Model" button. For instance, load the CI model for Smart Green House. Load the model. Choose the model. Click. You can then view the CI knowledge model then, in the Archived CI Models you can see your Smart Green House. Finally, every page features the "Download CI Model" button located at the buttom-right corner for downloading the model. Congratulations on successfully uploading and downloading the CI model using the QCI&AI-FML Learning Platform. See you next time. Bye-Bye.

Hello, my name is Mei-Hui Wang and I come from the National University of Tainan, Taiwan. Today, I will use Travel Recommendation System as an example to introduce and demonstrate how to connect human knowledge model to the machine model using the QCI&AI-FML learning platform and its associated learning tools based on the MQTT protocol so let's get started Go to Archived CI models and select the Travel Recommendation System. Then, click on the 'Conduct CI Inference' button. This will take you to this page. You can load the inference data or add data individually. For example, load the inference data with 3 records. Click on the 'MQTT Connection' button to activate the MQTT function. Choose a suitable MQTT server, click on mcutity connection button to choose the mcutity server for example the NUTN server, and set the topic for publishing such as 'QCIAI_MeiHui Please ensure you give a unique topic name to avoid the  QCI&AI-FML learning tool from receiving data not published by you Then, proceed to connect to the server. If successfully connected the icon color will turn green Next, connect the QCI&AI-FML Learning Tool to the MQTT Server. But, first connect the Learning Tool to your computer's USB port using the Thonny application. Then,open the python code const.py, to set the correct WiFi account and password. Set the valid MQTT username,  password, address, and port number. Then, set the choose the NUTN server as the MQTT server MQTT_AIFMLTOPIC to be the same as in the QCI&AI-FML Learning Platform to ensure the learning tools receive data from the platform.Following this please reset the learning tool before running the code. then open the mqtt travel recommendation file please make sure the machine knowledge matches the CI knowledge model before running the code. For example: NotRecommended is less than or equal to 35. Recommended is between 35 and 65. VeryRecommended is greater than 65. Now, let’s run the python code. Next, we infer the results for these three data When the learning tool receives this data it will respond in various ways such as displaying the information on the LCD verbalizing the information or triggering hardware actions like rotating a fan Let’s publish the result of the first data: 'If Expense is Low, Evaluation is Bad, and Distance is 200, then RecommendationLevel is NotRecommended' to the learning tool. Then, publish the first data. Then, publish the second data. Then, publish the third data. Finally, we already finishing connecting the CI knowledge model with the machine model using the QCI&AI-FML learning platform and learning tool Congratulations! See you next time. Bye-bye

Hello, my name is Mei-Hui Wang, and I come from the National University of Tainan, Taiwan. Today I will use Travel Recommendation System as an example to introduce and demonstrate how to make QCI Inference using QCI&AI-FML Learning Platform. So, let's get started. So, go to the home page and click on the MQTT button. This page offers CI Inference and QCI inference options. If the QCI&AI-FML Learning Tool is available, please set up MQTT connection carefully by assigning the correct names for topic subscribed and topic publishing. On the other hand, if the Learning Tool is not available, simply Activate MQTT service by providing arbitrary names. Assuming the QCI&AI-FML Learning Tool is available, please ensure these two topics match with "Topic Publishing" on the Learning Platform. This is crucial for the Learning Tool to receive human knowledge from the Learning Platform. This topic should match with the "Topic Subscribed" on the Learning Platform. This ensures that Learning Platform receives machine data from the learning tool. Once these settings are configured, proceed to activate the MQTT server. Next, click on the "Upload CI Model" button to upload your CI model to the QCI&AI-FML Learning Platform. You have the option to select either CI Inference or QCI Inference by choosing to send Test Data or Send Custom Data. It's important to note that QCI Inference only supports Mamdani Inference model. For instance, click on the "Send Custom Data" to obtain CI inference results and then publish it to the Learning Tool. Once published, the message will be displayed in the message box, and the Learning Tool will then response to it. If we choose "CI Inference" Send Custom Data the result is 50.000, and we choose "QCI Inference" Send Custom Data result in 49.773. These two values are very close. When selecting the QCI Inference and sending custom data Expense is 190, Evaluation is 33, and Distance is 23, Recommended, the Quantum Inference Model and two bar charts (count and probability) will be displayed here. You can save these three figures individually by right-clicking your mouse. When you send the data repeatedly, you will notice that the result very slightly each time. Recommended. Recommended. When the given data: Expense is 190, Evaluation is 35, Distance is 23 the result is 49.505. The membership degree for the linguistic terms are as follows: Therefore, the quantum register for Expense is here, for Evaluation is here, for Distance is here, after taking the square root of the membership degree. Check Rule No. 1 as example. If Expense is Low, Evaluation is Bad, Distance is Near, then the RecommendationLavel is NotRecommended. Expense_Low is 00 after mapping between binary string and fuzzy linguistic terms, so Expense0 and Expense1 have an X quantum gate.Total count for 001(NotRecommended) is 4060 and for 100(VeryRecommended) is 3940, after executing the circuit for 8000 shots. The probabilities for 001(NotRecommended) is 0.507 0.507 and for 100(VeryRecommended) is 0.492. These probabilities are then aggregated to the output fuzzy set, resulting in a RecommendedLevel of 49.505. Finally, we already finished QCI Inference Model using the QCI&AI-FML Learning Platform and Learning Tool. Congratulations!! See you next time. Bye-Bye.