Transfer of learning
Posted: Sat Dec 21, 2024 3:45 am
How to train models?
In order to classify objects, deep learning models need to be trained, and to achieve this there are three ways to do it, which we detail below:
From scratch
Although less common, it is a positive way to train in the sense that it works for novel applications or those that have a large number of output categories.
What is needed to do this is to collect a lot of labeled data while establishing a neural network dynamic in which the most relevant features are learned.
Its great benefit is that less data is recorded than in the previous case and the time is reduced from hours to minutes.
This is an adjustment to a model that had already been trained, that is, what has already been learned is transferred to it.
Therefore, the model is given new data that includes classes that would gambling email list have been previously unknown.
So transfer learning needs an interface with the internal factors of the pre-existing network, so that adjustment and improvement is possible in a very specific way and according to the new assignment.
Aspect Extraction
This process is also unusual, but it consists of extracting from the network the features already learned by the model and by the layers of neurons at any time during the training process.
These aspects or features can then be used as input to a machine learning model such as support vector machines (SVM).
Deep learning examples
To give you a clearer idea, here are the sectors in which deep learning models are used, which will help you imagine the magnitude of their benefits.
Defense sector
With the help of deep learning, objects can be identified in satellites, just as safe or unsafe areas can be detected for the benefit of troops.
Medicine
Deep learning applications are capable of detecting cancer cells automatically. With the help of a microscope, multiple multidimensional data are produced that can identify these cells and was developed by UCLA.
Autonomous driving
The automotive sector has also benefited from deep learning, as it is possible to automatically detect traffic lights, stop signs and even know if a pedestrian is crossing in order to avoid traffic accidents.
Industrial automation
In the industrial sector, it is possible to work with heavy machinery and detect whether there are people or objects nearby or areas that are not safe.
Electronics
Most people will recognize this example as devices that translate speech or users asking questions and having their questions answered.
Conclusion
Years, even decades ago, developers and experts hoped that one day they would be able to imitate the human brain with their machines almost perfectly.
In order to classify objects, deep learning models need to be trained, and to achieve this there are three ways to do it, which we detail below:
From scratch
Although less common, it is a positive way to train in the sense that it works for novel applications or those that have a large number of output categories.
What is needed to do this is to collect a lot of labeled data while establishing a neural network dynamic in which the most relevant features are learned.
Its great benefit is that less data is recorded than in the previous case and the time is reduced from hours to minutes.
This is an adjustment to a model that had already been trained, that is, what has already been learned is transferred to it.
Therefore, the model is given new data that includes classes that would gambling email list have been previously unknown.
So transfer learning needs an interface with the internal factors of the pre-existing network, so that adjustment and improvement is possible in a very specific way and according to the new assignment.
Aspect Extraction
This process is also unusual, but it consists of extracting from the network the features already learned by the model and by the layers of neurons at any time during the training process.
These aspects or features can then be used as input to a machine learning model such as support vector machines (SVM).
Deep learning examples
To give you a clearer idea, here are the sectors in which deep learning models are used, which will help you imagine the magnitude of their benefits.
Defense sector
With the help of deep learning, objects can be identified in satellites, just as safe or unsafe areas can be detected for the benefit of troops.
Medicine
Deep learning applications are capable of detecting cancer cells automatically. With the help of a microscope, multiple multidimensional data are produced that can identify these cells and was developed by UCLA.
Autonomous driving
The automotive sector has also benefited from deep learning, as it is possible to automatically detect traffic lights, stop signs and even know if a pedestrian is crossing in order to avoid traffic accidents.
Industrial automation
In the industrial sector, it is possible to work with heavy machinery and detect whether there are people or objects nearby or areas that are not safe.
Electronics
Most people will recognize this example as devices that translate speech or users asking questions and having their questions answered.
Conclusion
Years, even decades ago, developers and experts hoped that one day they would be able to imitate the human brain with their machines almost perfectly.