Arthur Samuel, who is considered as a pioneer of machine learning defined machine learning as “Field of study that gives computers the ability to learn without being explicitly programmed.
However, it was Tom Mitchell, an American computer scientist having PHD from Stanford who defined Machine learning as
“A computer program is said to learn from experience E with respect to some task T and some performance measure P, if its performance on T, as measured by P, improves with experience E.
Machine learning is basically a process in which algorithms are used to train the computers in learning about a problem. In one of our boot camps, we demonstrated how humans learn and how we use the same process for machines to learn. We had a Chinese character data set, with the character written in different styles and handwriting,we asked our students to look at that character very carefully and label that character in their brain. Once the character recognition was over, we handed them a piece of Chinese text and asked them to recognise the character they just trained their brain on. The audience was able to recognise the character they just learned. Human brain capability to learn is replicated on machines but with a higher computing power.
Machine learning is categorised mainly in two forms
In this type of Machine learning, the program is provided a data set, which we call it as a training data ,which is labelled and has an outcome associated with it. After learning from this training data set ,the algorithm applies a function to predict the values for the test data ,which is provided for computation. There are multiple models of Supervised form of learning. In this supervised form of learning the input as well as the desired output is present. These are the training examples which we provide it. For example, when we are provided a character of a language, we also have the information regarding the pronunciation of that character which we call as an output. Once the Machine is trained to understand the training examples ,when it encounters a test character , it knows what the pronunciation of that character needs to be and then the algorithm will provide the output as desired. There are possibilities that the test predication can vary from the level of the confidence, for example, if the character for recognition is not much understood by the machine learning algorithm it gives a certain degree of error and provides the confidence level of possible understanding and match.
This type of machine learning is used when we do not have a labelled data and we have no information about the data classification. The machine learning algorithms then cluster the data into different structures. This means that the machine learns on its own and groups the data into clusters which can then be visualised to understand the type of data we have and the information that we can derive from it.
How Machine Learning is contributing to Artificial Intelligence applications
- Disease Diagnosis and medication: – Diagnosis of the X Rays and the MRI scans.
- DNA based personalised medicines to treat the life-threatening diseases and extend the human life.
- Mental Health Support: – Algorithms that highlight the emotional state of a human and risk management based on the context and the speech. Apps like Moodpath , Talklife, Youper, Calm Harm.
- Mobility – Autonomous vehicles has the potential to safeguard the traffic and completely make it accident free and it may also allow the senior citizens and the disables to use the vehicles to travel.
- Education – Each child has his/her different needs and the brain’s ability differs from a child to another child. The next generation of the education will be AI based which will eventually help personalise the education and focus on the child’s ability to respond.
- Virtual Chat-bot mentors which can guide to complete the problem. It can search over vast spheres of educational material.