Types of Parameters in machine learning So, in machine learning, we often deal with two types of parameters. Okay, so one type of parameter is called the model parameters, and the other type is the hyperparameters. So, let's try to understand each of these types of parameters with the definitions. Model parameters are those parameters of the model that can be determined by training with the training data, and these can be considered as internal parameters. Okay, so we know that in machine learning, we often deal with data and we have some model. You can think about the models like logistic regression, support vector machines, or k-nearest neighbors, and we feed the data to the model. What does this model try to do? It will try to find the most optimum parameters. So, these parameters that are determined by the model using the data are called the model parameters. Once we carry out this training process, the model will try to find the best parameters. Examples of these model parameters are weights and bias. Okay, so if you consider a linear regression model, we have the equation as \(y = wx + b\), where \(y\) is my dependent variable (or you can call this the target), and \(x\) is my independent variable. So, if we consider an example, let's say that we want to determine what salary a person can make depending on years of work experience. In that case, \(y\) will be the salary that we are going to predict, and \(x\) will be the number of years of experience. We know that the equation of the line is given by \(y = mx + b\), and here, \(w\) is similar to the slope. So, we know that we represent the slope as \(m\) and the intercept as \(c\). Instead of that, we can write it as \(w\) and \(b\). So, \(w\) represents your weight, which is similar to the slope of a line, and \(b\) represents your bias, which is again similar to the intercept of the line. These \(w\) and \(b\) will be your model parameters. Okay, so we initiate with a random weight value and bias value, and once the model is fit with the data, it will try to find which is the best parameter value for this \(w\) and \(b\). These set of parameters are called the model parameters. Now, there are another set of parameters called hyperparameters. Hyperparameters are the parameters whose values control the learning process. These are adjustable parameters used to obtain an optimal model. These parameters are also called external parameters. Let us try to understand these hyperparameters. As I have told you, these set of parameters, weights and bias, which we call as model parameters, are determined, or these values are derived based on the dataset. But the hyperparameters are something that we give to the model. So, this hyperparameter determines how your model is going to perform. Okay, so this set of values are determined by working on the data set, whereas this is given to the model and these parameters, you know, dictates how a model is going to train from that particular dataset. So hence, we call these as external parameters because we can control these parameters. Okay, so you can change the values of hyperparameters, but you cannot do it directly when it comes to the weights and bias, which we call as model parameters, because the model tries to find the best parameters on its own, whereas we control the hyperparameters. Okay, and the examples of hyperparameters are learning rate, number of epochs, and n estimators. So, these are a few examples. Again, there are a lot more examples of hyperparameters. So, learning rate is nothing but how much do you want to change your weight and bias. So, if you have seen my gradient descent videos before, in that we have seen that when a model is training on a dataset, what it will try to do is initially, it will, you know, randomly assign some values to this weight and bias. So, we usually initiate with, you know, a value as zero. So, after that, it will try to improve the performance of the model. So, in other words, it will try to minimize the loss function or the cost function, and in each step, it will try to change the value of weight and bias and again calculate the loss function for that particular weight and bias. So, in this case, the learning rate determines how much change you want to give to your weight value and bias value. So, this is your learning rate. So, that is an example of learning rate. And then we have the number of epochs. You can also call this the number of iterations. This determines how many times, you know, your model is going to go through your data. So, if you represent this number of iterations as 100, it means your model will go through the data 100 different times. You can give it a thousand or whatever you want. So, it all depends on how efficient you want your model to be. And when it comes to a random forest model, there is a hyperparameter called *n_estimators*. So, an estimator is nothing but the number of decision trees that you want in your random forest. I hope you know that random forest is nothing but an ensemble model where we have several decision trees. So, it is a group of decision trees, and this *n_estimators* is nothing but the number of decision trees we want in our random forest model. Okay, so these are the two sets of parameters: one is the model parameters, and the other set of parameters is your hyperparameters. The most important thing that you need to remember is that model parameters are derived by the machine learning model by going through the data, by understanding the data, whereas these hyperparameters are going to determine how your model is going to train. So, these sets of parameters we can change depending on our requirement.