Multiple Definitions of Machine Learning
There are multiple definitions on the internet about machine learning. But, here we will look at two most important machine learning definitions which are comprehensive and which give you an insight about you should expect in coming lessons:
- Arthur Samuel ( 1959 ) – Field of study that gives computers the ability to learning without being explicitly programmed.
- Tom Mitchell ( 1998 ) – 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 measure by P, improves with experience E.
DATA –> ALGORITHMS –> LEARNING –> MAGIC
Arthur Samuel was the first person who originally coined the term ” Machine Learning” and it is due to his contributions that it is a specialized field today growing at a rapid pace. In act, it is growing at such a blistering pace that some of us are insecure whether it will replace our jobs and make us irrelevant in the near future.
Basic Mathematics & fairly good knowledge of Statistics is essential to get started with Machine learning and its algorithms. Also, implementation of these algorithms requires a programming language like Python, R, or in our case Kotlin.
A key feature of machine learning is the concept of “self-learning“. This means the application based on models developed by statistical analysis to detect patterns in the data improves the performance of the model by the feedback of information without explicitly telling the program what to perform.
Machine Learning involves data input to make decisions based on a model. These decisions are the output of computations based upon relationships and patterns between the various data points. The input data is the most significant cog of the wheel which determines the output without any human interference or bias. The human touch is needed to fine tune the parameters and help develop a conducive environment to get desirable results. The machine learning engineer is responsible for feeding the model with clean and relevant data otherwise arbitrary results will spoil the output accuracy.
Just like humans learn each and every day, from their experiences and adapt to life, learning from experiences and taking better decisions in future. In a similar manner, the consequential feature of machine learning is that it learns from experience to give better predictions and outcomes. As mentioned above, the input data is the single-most important factor to produce desirable outcomes and accurate predictions. Half-cooked input with insufficient coherence will under-perform from its potential and give low quality results.
Computer programmers make rigid rules in their code to produce expected output but machine learning algorithms can never be restricted by rigid rules otherwise a random data point can destroy the whole model as flexibility is not provided for learning from distinct data points to give desired outcomes. But, here we should not be confused that more input data will solve the problem of giving low quality data to the model, more data will only increase processing time and waste processor cycles leading to loss of time and money.
Training and Testing Data
In Machine Learning, the input data is divided into 2 parts:
- Training Data
- Test Data
The data that is used to develop the model is called Training Data.
When the model has been developed to the level of satisfaction, then the remaining data is split to produce Test Data. This data is then fed to the model to check the efficacy of the model and the accuracy of predictions.
Exploration of Machine Learning and related branches
Let’s learn about machine learning is connected to the other blocks in computer science and mathematics. Data mining, artificial intelligence, deep learning and computer programming are all branches of computer science which includes anything and everything related to computers. Data science has a separate branch called Data Science which includes methods and systems to develop models and algorithms using which we can get insights about data using computational power.
Artificial Intelligence or AI is the ability of a machine to perform intelligent cognitive tasks. Its in its nascent stage at the moment because of lack of skilled professionals in the domain. It has immense potential to grow and is a niche subcategory of computer science which will engulf us in all walks of life. Machine learning is a prerequisite for AI as many algorithms can be applied in AI sub-fields like Natural Language Processing and Image/Audio Analysis.
Data mining is another fields closely related to Machine Learning as it is used to extract information from huge amounts of raw data. Pattern recognition and probabilistic models are used in data mining to predict accurate analysis of data.