(Not a big insight, but anyway))Disciplines which are in the heart of Data Science:
- Statistics and Statistical learning theory
- Linear Algebra (LA)
- Algorithms and data structures
- Optimization
They form a theoretical background necessary to understand how most of the Machine Learning/Data Mining algorithms work. Why do you need to know how internals work? Because most of the ML algorithms are "leaky abstractions" [1],[2].
The Data Mining is a process of finding the patterns in the data. LA is not only the tool, but it inherently contains elements of pattern recognition. For example the process of matrix factorization reveals patterns in the column / row space of the matrix.
The Data Mining is a process of finding the patterns in the data. LA is not only the tool, but it inherently contains elements of pattern recognition. For example the process of matrix factorization reveals patterns in the column / row space of the matrix.
- [1] "Yes you should understand backprop" by Andrej Karpathy
- [2] HN discussion: "Backpropagation is a leaky abstraction"
No comments:
Post a Comment