It is going to be interesting to see how society deals with AI, but it will definitely be cool-Colin Angle.
To know about Machine Learning one should know what is Artificial Intelligence. Artificial Intelligence is a Science and Engineering to make machines smart and Intelligent especially developing intelligent programs. It is coined by famous personality John MacCarthy and is called as one of the founding fathers of AI. It is not a new word but is reflected back in 1955 by John.
Machine Learning is an application of AI that makes the system to learn from experience not by explicitly programmed. It just works on the same basics as human intelligence works i.e learning from experience. Just imagine the magic that machines are automatically identifying the unlabelled data. Training is an important phase of Machine Learning and once training is complete machines start behaving like the human.
Machine Learning Algorithms: Machine Learning algorithms are either supervised, unsupervised, semi-supervised and reinforcement.
1) Supervised Learning: It can be compared with a teacher i.e for every input there is a corresponding target. If the target is in the form of some classes then it is called classification problem. If the target is continuous it is called regression problem. ‘
2) Unsupervised Learning: It is without a teacher where only a set of inputs are present. This algorithm tries to find the relationship between inputs more logically. Unsupervised learning refers to clustering which will create difference clusters and will fit the new data in the appropriate cluster. Other than clustering Anomaly Detection, Hebbian Learning, Latent Variable model such as expectation maximization algorithm, Blind Signal separation, PCA and singular value decomposition can be done.
3) Semi-Supervised: It falls between supervised and unsupervised both which uses labeled and unlabeled data for training. Basically, it is a hybrid version of supervised and semi-supervised to improve learning accuracy.
4) Reinforcement Machine Learning: It is close to human learning. Basically, these are the algorithms by which machines are trained to act in a given environment.
In simple words, every action has some impact on the environment and this environment provides rewards which help in learning. Basically, it is also called as learning through trial and error interaction with the dynamic environment. Feedback is provided that evaluates the learner performer. Reinforcement means reward from the environment.
Basic steps involved in Machine Learning:
1) Gathering Data: First and most important step in machine learning is together the appropriate data. the quality and quantity of data collection is crucial while building the model.
2) Data Analysis and Pre-Processing: Once you are done with data gathering look carefully at the data. If there is no target, unsupervised learning algorithm should be clicked in our mind and if there is specified target, supervised learning must be clicked in our mind. This is the job of data analysis. Once analysis is done it is time to perform cleaning and pre-processing which include removing undefined values from data and operations like image enhancement, skewing, histogram equalisation and sharpening in image processing.
3) Feature Extraction and Feature Selection: Once data or images are pre-processed then the time of feature extraction and selection comes. At the first stage features are extracted using suitable feature extraction algorithm like principal component analysis, partial least square, auto-encoders and then features are selected using appropriate algorithm like one may opt for Pearson’s co-relation, Anova, LDA and chi-square.
4) Classification or Regression Model: After feature selection, it is time to train the model. Training algorithm depends upon whether the data is continuous or having classes or categories. Hence search for either classification model or regression model. Further to improve the accuracy these models can be tuned with best parameters like tuning of SVM using Grid Search.
Pictorial representation of steps are shown in the following image.
Growth of Machine Learning: Machine Learning is an integral part and preferable approach of following:
- Speech Recognition
- NLP (Natural Language Processing)
- Computer Vision
- Weather Forecasting
- Stock Prediction
- Face Recognition
- Age Invariant
Applications of Machine Leaning: Following are some applications of Machine Learning.
Different Disease Prediction, Image Classification, Object Detection, Fingerprint Matching, Robot Building, Self Driving Cars, Offline Handwritten Optical Character Recognition and many more. It is utilising in most of all fields so there are a lot of applications of Machine Learning.
Conclusion: From all of above discussion, and observations it can be concluded that future belongs to AI and more precisely to Machine Learning. Seeing its application in forensic, healthcare, mining, education, assisting human, Stock Prediction and future forecasting. It is quite obvious and necessary to learn it. Even big companies like Facebook, IBM, Amazon, Google, Microsoft has entirely shifted on it because of its magical results and positive services to mankind. Hope this article is helpful somehow. I have tried to introduce you with Machine Learning and AI.