Big Data, Machine Learning and Deep Learning
Big data is large volume data collecting by 4 manner:
- buying
- questionnaire
- internal data
- free data from organisation or government after collected the big data, it will make use of analytics.
Regarding the purpose of the analytics, there are 3 categories.
- acquire the human habits
- predict the future trend
- prevent crime and arrest the suspect Data category can be divided to CCTV record, POS data and scraping form website.
in general, we have 2 years data. Firstly, we use 1.5 years data to training. After training, we satisfy the result. We will take the rest of the 0.5 year data to run the training program in order to find out the training program whether it is satisfied or not.
Machine Learning can be identify to 3 type:
supervisor: give data and answer unsupervisor: give data but no answer reinforcement: environment only for example, gaming AI, chess AI
Different between Machine Learning and Deep Learning
Machine Learning is significantly supported by big data. As input from big data, Machine Learning specify the data of identification one by one from human hand. All feature(data) defined by human in details, but Deep learning do not do so.
For instance, present one photo to the machine in order to identify the object belong to cat or dog.
Machine Learning need man power to classify huge photo(big data) to tell the machine which photo whether they are cat or dog. Human must identify a number of sample photo(data) of feature like the outlook of face, eye edge, mouth edge and nose edge to machine. If outlook of face is square, inside the square have two circle(eye), one triangle(nose) and small oval(mouth), this photo of feature can tell to machine it is cat.
But Deep learning can find the edge by itself. if machine find the object face edge, then find the face have two circle(eye), one triangle(nose) and one oval(mouth), this photo of feature can be analyzed to a cat. If the mouth is rectangle, the machine can tell human that is dog.
Machine Learning need a lot of human power but the processing time will be short. Deep Learning need lack of human power but the processing time is much long.
Tags: #big data