Table of Contents
Introduction
As a field of computer science, machine learning enables computers to learn without the need of being programmed. It is useful in various computing tasks in which programming, designing explicit algorithms, ensuring good performance is infeasible. This method is used to devise difficult algorithms and models which can lead themselves to prediction referred to as predictive analytics. Such models enable researchers, engineers, data scientists, and analysts to offer repeatable and reliable decisions as well as uncover hidden insights by leaning from trends and historical relationships. As time goes by, people have transformed from the use of traditional learning techniques to computerized systems where the computer has become among the major resources in learning. The more knowledge one gains on how to use a machine is the more they are able to utilize the resource. This paper explores the various areas and issues relating to machine learning.
The principles behind the field of machine learning
Three main principles are used in the field of machine learning which are representation, evaluation, and optimization. The three principles are meant to help towards meeting the machine learning objectives. Starting with representation as the number one principle, it focuses on teaching the learner to understand how to represent knowledge in an appropriate manner. Such models and techniques used in enabling or enhancing good representation include the decision tree diagrams, graphical models, sets of rules, neural networks, model assembles, instances, and support vector machines (Brownlee, 2015).
The other principle is evaluation. This principle is concerned with the way in which program candidates should be evaluated. Such include recall and predictions, accuracy, likelihood, squared error, cost margin, posterior probability and entropy K-L divergence. Machine learning calls for proper evaluation to ensure that every decision made is accurate and the most optimal decision for the specific situation. In optimization principle, machine learning focuses on making the maximum use of every opportunity or resources that an individual comes across. This is mainly concerned with the search process which is the way in which candidate programs are generated. Optimization entails; combinational optimization, constrained optimization, and convex optimization.
Prerequisites needed to enable machine learning
Basic linear algebra
Before starting to study machine learning, it is important to have some basic knowledge in liner algebra. Linear algebra equips the learner with some basic knowledge in terms of the connection between linear transformation, matrices, and linear equation systems. Such knowledge is important for the leaner to understand the connection between the principles of machine learning and its functions. Without the basic linear algebra knowledge, it becomes difficult for one to succeed in machine learning (Singh, 2014).
The other important prerequisite for machine learning is programming experience. The learner should have some experience in programming which entails the key concepts of computer science. Understanding the key concepts to computer science is important before one goes to the study of machine learning. These concepts are highly utilized in the study of machine learning and without some experience in programming; it becomes difficult for the learner to succeed. Having some basic knowledge in probability and statistics is also important. Machine learning involves much use of statistics and probability. It is essential for anyone who study machine learning to have sufficient knowledge of statistics and probability. Finally, when studying machine learning, one should be in a position to do a lot of calculus which makes basic calculus skills another prerequisite for machine learning (Singh, 2014).
What breakthrough has occurred in the field of machine learning?
Scientific research team has recently come up with a new method of machine for predicting molecular behavior. This is a breakthrough that can help in the pharmaceuticals development and the new molecules design. As a result, it will be useful in enhancing the performance of solar cells, digital displays and the emerging battery technologies. The pattern identification in molecular behavior builds more knowledge in atomic interactions in molecules which then draws on the information for predicting new phenomena (New York University, 2017). At the same time, the research team came up with a machine that is useful for learning complex interatomic interactions normally prescribed by mechanical calculations without necessarily performing the intricate calculations. In constructing the machine, a small sample of molecule was selected for training the algorithm to stimulate the complex chemical behavior in the molecule.
Challenges that remain in the field of machine learning
One of the major challenges that still persist in the field of learning is the use of kernel methods. Presently, most learning algorithms work on fixed and flat feature vectors. It is important for every new type of data to have new learning algorithms. However, a particular data interface is used for several learning methods while a particular method can be used for multiple types of data. The other challenge is that most of the machine learning focuses on regression and classification which to some extent, has changed from the use of discriminative methods to more general techniques. Notably, discriminative methods are known to outperform generative methods. The change from using discriminative methods to generative methods is, therefore, a challenge to the field of learning. Lastly, is concerning the use of interpretable methods which have resulted into lack of insight and trust in the rule (Ho & Jaist, 2017).
How machine learning will impact my major study
Preventing Fraud
As a financial service provider, I will have a greater responsibility other than that of protecting clients against fraudulent activity. In every year, about $50 billion is spent on financial fraud in American. The traditional means of protecting clients’ information are no longer secure. As a result, criminals are taking advantage of the challenge and hence there is a need for better ways of protecting clients’ information as I become a financial service provider. In order to stay a step ahead of hackers, machine learning will be extremely useful. With machine learning, I will be able to outthink the criminals and hence enabling the applications to prevent security breaches (Ternovyi, 2017).
Risk Management
Unlike the traditional methods where creditworthiness is predicted using static data from financial reports and loan submissions, the machine learning skill analyzes the individual’s financial position based on the current market trends. The new technology detects rogue investors who work in unions across several accounts which would be impossible or a difficult task for the investment manager. At the same time, machine learning enhances effectiveness and efficiency in finance. Through the machine learning platform, investment managers are able to concentrate more on productive activities like serving customers (Ternovyi, 2017)
Investment Prediction
When using the computer aided services, an investor can place an order once a stock reaches some predetermined price as well as to sell the stock if the price falls to some particular limit. Automatic functions make such platforms to trade easier both for small and large investors. While investors can make recommendations as a result of the market automated analysis, they also face some limitations which can all be solved through using the automated functions. Ternovyi (2017) provides that hedge funds have recently changed from the traditional methods of analysis to machine learning algorithms in predicting fund trends. Through the use of machine learning, fraud managers can identity any changes in the market earlier than they could do with the traditional investment models.
Conclusion
According to the above explanations, machine learning is beneficial in the various areas of study. Currently, individuals are changing from using their old systems as they take advantage of the computerized functions. Among the basic principles of machine learning are representation, evaluation, and optimization. However, machine learning is never easy and some factors must be considered before someone enrolls to pursue the course. Some of the prerequisites are programming experience, basic linear algebra, and knowledge in calculus. In pursuing my finance career, machine learning will be helpful in preventing fraud, performing risk management, and investment prediction which will be part of my responsibility as a financial service provider. With the knowledge of machine learning, I am confident that I will be in a position to perform my duties and responsibilities better than I would do through the traditional methods and hence a higher level of success.
- Brownlee, J. (2015). Basic Concepts in Machine Learning. Retrieved from: https://machinelearningmastery.com/basic-concepts-in-machine-learning/.
- Ho, B. & Jaist, (2017). Challenges in Machine Learning and Data Mining. Retrieved from: https://www.jaist.ac.jp/~bao/Writings/Challenges%20in%20ML&DM.pdf.
- New York University. (2017). Scientists develop machine-learning method to predict the behavior of molecules. Retrieved from: https://phys.org/news/2017-10-scientists-machine-learning-method-behavior-molecules.html.
- Singh, J. (2014). What are prerequisites to start learning machine learning? Retrieved from: https://www.quora.com/What-are-prerequisites-to-start-learning-machine-learning.
- Ternovyi, D. (2017). 7 ways Fintechs use machine learning to outsmart the competition. Retrieved from: https://igniteoutsourcing.com/publications/machine-learning-in-finance/