Humans learn from past experience and machines follow instructions given by humans. Humans can train the machines to learn from the past data. Machine learning is a lot more than just learning, it is also about understanding and reasoning.
Machine learning is the science of making computers to learn and act like humans by feeding data and information without being programmed in a clear and detailed manner.
By looking at the diagram below, there is an ordinary system and with the help of machine learning, system will take the data and will learn from it. After learning, it will predict the output. Biggest part of machine learning is, the system will improve from the prediction and will find a new solution.
 |
| Revolution of Machine Learning |
Different stages of machine learning can be seen in the diagram below, first there is a need to define the objective, very important to know what one is expecting from the model to predict. Once the objective has been defined, there is a need to collect the right data and to prepare it (right data in, right answer/data out). After the preparation phase, there is a step of selecting an algorithm. After selecting an algorithm, model will be trained and tested to predict the outcome of the data used for the model. Based on training and testing phase, prediction comes out and then the model can be deployed.
 |
| Stages of Machine Learning Model |
In modern world. machine learning is considered to be one of the rapidly growing field of computer science having important and widely applicable effects in applications like agriculture, speech recognition, telecommunication, computer networks, banking and medical diagnosis etc. Although being rapidly growing field, programs associated with machine learning are sometimes unsuccessful to deliver expected results. Some of the reasons for being unsuccessful to deliver expected outcomes includes absence of suitable and access to data, problems related to privacy, imperfectly chosen machine learning algorithms and tasks, incorrect tools and inexperience people, shortage of resources and evaluation problems. Some famous failed models of machine learning includes, a self driving car (2018) from Uber that was unsuccessful to detect a pedestrian, who was killed after a collision. Several attempts were made by IBM Watson to use machine learning in healthcare failed enormously to deliver even after making billions of investment.
Machine learning make adequate preparations for fully automated methods that can be used to identify the existence of different patterns in big data and afterwards use them to accomplish different tasks. There are four types of machine learning categories that can be taken into account:
- Supervised Learning
- Unsupervised Learning
- Semi-Supervised Learning
- Reinforcement Learning
There are certain approaches available to solve these machine learning categories which includes linear models, state vector machines, decision trees, Gaussian processes, deep learning etc. Recently machine learning has shown growth due to significant advancements in deep learning considering learning from data (highly dimensional) for example images, videos and time series data. Usage of GPUs making the computational power to escalate, advances in software and libraries like pycharm, tensorflow, caffe have been the reason for the improvement in machine learning.
Recommended Links to get deep into machine learning:
- https://www.slideshare.net/Simplilearn/machine-learning-tutorial-part-1-machine-learning-tutorial-for-beginners-part-1-simplilearn/Simplilearn/machine-learning-tutorial-part-1-machine-learning-tutorial-for-beginners-part-1-simplilearn
- https://web.archive.org/web/20170320225010/https://www.bloomberg.com/news/articles/2016-11-10/why-machine-learning-models-often-fail-to-learn-quicktake-q-a
- https://www.kdnuggets.com/2018/07/why-machine-learning-project-fail.html
- https://www.economist.com/the-economist-explains/2018/05/29/why-ubers-self-driving-car-killed-a-pedestrian
- https://www.wsj.com/articles/ibm-bet-billions-that-watson-could-improve-cancer-treatment-it-hasnt-worked-1533961147
- http://cds.cern.ch/record/998831
- https://arxiv.org/abs/1603.04467
- https://link.springer.com/article/10.1007%2Fs10994-006-6226-1
- https://www.sciencedirect.com/science/article/abs/pii/S0893608014002135
- http://profsite.um.ac.ir/~monsefi/machine-learning/pdf/Machine-Learning-Tom-Mitchell.pdf
