Super model parameters and model parameters in machine learning are different in terms of the role, and other sources, and ultra-parameter model is often referred to as model parameters so that it can be confusing for beginners. This article gives the definition of the model parameters and the model parameters over, and compared, the differences noted on both the nature: the model parameters are arranged inside the model variables, the model parameters can be estimated data; super model is a model parameter external configuration must manually set this parameter.
When we do research, we will encounter many of the terms. Sometimes, the term of the same name will appear in different areas of research. For example, statistics, "model parameters" is often used in economics and the "super model parameters" in machine learning also exist.
In the field of machine learning "model parameters", "super-model parameters" in the role, and other sources are different, beginners without a clear understanding of the two, tend to learn it would be more difficult, especially those from statistical and the field of economics beginners.
In order to let everyone in the application of machine learning, "model parameters" and "ultra-parameter model" There is a clearly defined, in this article, we will discuss specific terms. If you find this article also looks a little bit hard, or want to learn the system of artificial intelligence, it is recommended that you go to bed long artificial intelligence tutorial. For the great big God, tutorials not only easy to understand, and very humorous. Click here you can view the tutorial.
First, let's look at the "parameters" What is?
As part of the model parameters from training data learned from history is the key to machine learning algorithms .
Statistics in the "parameters":
In statistics, you can assume that the distribution of a variable, such as a Gaussian distribution. Two parameters of the Gaussian distribution are the mean ([mu]) and the standard deviation (sigma). This is effective in machine learning, where these parameters can be estimated from the data and used as part of the prediction model.
Programmed in the "parameters":
You may pass parameters to the function programming. In this case, the parameter is a function parameter, there may be a range of values. In machine learning, the specific model you are using is the function, the need for new data parameters to predict.
"Parameters" and "model" What is the relationship?
According to the classical machine learning literature may be considered as the model assumptions, the parameter is adjusted depending on the particular data set of assumptions.
Whether the model has a fixed or variable number of parameters, the model determines a "parameter" model or "non-reference" model.
What is the model parameters?
In simple terms, the parameters of the model is the internal model configuration variables, it is possible to estimate the value of the data.
Specifically, the model parameter has the following characteristics:
We need to model parameters when the model predictions.
Model parameter values can be defined model function.
Learning model parameter estimation or data obtained by the data.
Model parameters is generally not practical to manually set.
Model parameters are usually stored as part of the learning model.
Commonly used optimization algorithm to estimate model parameters, optimization algorithm is an effective search for the possible values of the parameters carried out.
Some examples of model parameters include:
Artificial neural networks weights.
Support Vector Support Vector Machine.
Regression coefficient of a linear regression or logic.
What is the super model parameters?
Super external model is a model parameter configuration, its value can not be estimated from the data obtained.
Specific features include:
Super model parameters often used to estimate model parameters of the process.
Super model parameters are usually specified directly by the practitioner.
Super model parameters can usually use heuristic methods to set.
Super model parameters are usually adjusted in accordance with a given predictive modeling problem.
How to get its optimum value: For a given problem, we do not know the optimal value super model parameters. But we can use the rule of thumb to explore its optimal value, or copy the values for other problems, can also be a method by trial and error.
Some examples of super-model parameters include:
Training and learning rate of neural networks.
Support for C and sigma ultra vector machines.
k k neighborhood.
"Model parameters" and "super-model parameters."
Link between the two:
When adjusting the machine learning algorithms for specific problems, such as when using a grid search or random search, you will adjust the model or the super command parameters to find a model can predict the most skilled model parameters. Many important parameter in the model can not be estimated directly from the data obtained. For example, in a K ... This type of model is referred to as model parameters adjustment parameter, because the analysis is not available to the formula for calculating an appropriate value.
- pp. 64-65, use predictive modeling , 2013
distinguish:
Super model parameters commonly referred to as model parameters, this name is easy to misunderstand. A good rule of thumb to solve this problem as follows: If you have to manually specify a "model parameters", then it could be a super model parameters.
Further reading
Hyperparameter - Wikipedia - https://en.wikipedia.org/wiki/Hyperparameter
What is machine learning hyperparameter? Quora - https://www.quora.com/What-are-hyperparameters-in-machine-learning
Super model parameters and model parameters What is the difference? StackExchange- https://datascience.stackexchange.com/qu
What is hyper-parameters? Reddit - https://www.reddit.com/r/MachineLearning/comments/40tfc4/what_is_considered_a_hyperparameter/
to sum up
After reading this article can be clearly defined and understood the difference between the model parameters and super model parameters.
In summary, the model parameters are estimated from data automatically, and super model parameter set manually, and the process for estimating model parameters.