[PRML] Ch 1. Introduction
업데이트:
This post is summary of the book Pattern Recognition and Machine Learning by Christopher M. Bishop
1. Introduction
- pattern recognition is automatic discovery of regularities in data through computer algorithms
- we can perform tasks such as classification using these regularities found in the data
- Machine Learning can be used to find the digits of these handwritten digits
- Though it could be tackled using handcrafted rules or heuristics, machine learning can give better results due to high variety of the data
Machine Learning
- Machine Learning is the process of finding a functions $y(x)$ which takes new digit image $x$ and generates an output vector $y$, which has same format as the target vectors
- because the training data only contains a part of all possible input vector, generalization is central goal in pattern recognition
- pre-processing is done to make pattern recognition problem easier
- it is sometimes called feature extraction
- test data must be pre-processed using the same steps as the training data
- pre-processing might be done to spped up computation
- pre-processing must be done in a way it doesn’t discard important information that might affect the overall accuracy
Types of Machine Learning
- Supervised Learning: training data is comprised of input vectors and corresponding target vectors
- Classification: target vectors are consisted of finite number of discrete categories
- Regression: target vectors are consisted of one or more continuous variable
- Unsupervised Learning: training data comprised of input vectors without any corresponding target values
- Clustering: disvocering groups of similar exampels within the data
- Density reduction: determining the distribution of data within the input space
- Visualization: projecting the data from a high-dimensional space down to two or three dimensions
- Reinforement Learning: the problem of finding suitable actions to take in given situation in order to maximize a reward
- there is a sequence of states and actions which has to be discovered by trial and error
- in many cases, the current action not only affects the immediate reward but also has an impact on the reward at all subsequent time steps
댓글남기기