Hi Guo,
In the provided code and context, Fis and chkFis are related to the process of training an Adaptive Neuro-Fuzzy Inference System (ANFIS) using MATLAB's Fuzzy Logic Toolbox. The relationship between Fis and chkFis and how training and checking data are handled during each epoch are crucial to understanding how ANFIS models are trained and validated to prevent overfitting. Let's break down your questions:
Relationship between Fis and chkFis
- Fis represents the Fuzzy Inference System (FIS) model that is being trained using the training data (trnData). It starts as an initial model generated by genfis1 from the training data and is then refined through the training process using the anfis function.
- chkFis is essentially a version of Fis that has been adjusted during training using the checking data (chkData). The purpose of chkFis is to monitor the training process for signs of overfitting. During training, ANFIS uses the checking data to validate the model at each epoch. If the model's performance on the checking data starts to degrade (indicating potential overfitting to the training data), ANFIS can stop adjusting the model or take other actions to prevent overfitting.
How Training Data and Checking Data are Handled
- During each epoch of training, ANFIS adjusts the FIS model (Fis) based on the training data to minimize the error between the model's outputs and the actual outputs in the training data.
- Simultaneously, ANFIS evaluates the adjusted FIS model on the checking data to monitor the model's performance on data it hasn't been trained on. This is done to ensure that the model generalizes well and isn't just memorizing the training data (overfitting).
- The checking is conducted after each training epoch, and adjustments to prevent overfitting (like stopping the training early or adjusting the model parameters) are based on the model's performance on the checking data.
Which Model to Use for Further Testing (Fis or chkFis)
- If your goal is to have a model that generalizes well to new, unseen data, you should use chkFis for further testing. This model has been adjusted to prevent overfitting based on its performance on the checking data.
- Fis represents the model trained purely on the training data without adjustments for overfitting. It might perform better on the training data but could perform worse on new, unseen data compared to chkFis.
In summary, chkFis is typically the preferable model for testing and deploying in real-world scenarios because it is adjusted to mitigate overfitting, ensuring better generalization to new data.
