Introduction
machine learning: look for a function to do something like speech recognition etc.
Learning Map
overview
no. | scenario | task/diff | difference |
---|---|---|---|
1 | Supervised Learning | labelled | |
2 | Regression | $f$ output: scalar | |
3 | Classification | $f$ output: classification | |
4 | Structured Learning | $f$ output: structure | |
5 | Semi-Supervised Learning | data is related to task | labelled & unlabelled |
6 | Transfer Learning | data is unrelated | labelled & unlabelled |
7 | Unsupervised Learning | Like word2vec | machine learns meaning |
8 | Reinforcement Learning | have no certain or unique ans | learn from critics, better or worse |
Supervised Learning
1. Regression
the output of target function $f$ is scalar.
2. classification
binary classification: the output of target function $f$ is yes/no.
Multi-class classification: the ouput of target function $f$ is classification
1&2 method: linear model and Non-linear Model, deep learning is one of non-linear model for complicated problems.
3. Structured Learning
the output of target function $f$ is structured data.
比如,根据一段语音得到对应文本。
Semi-Supervised Learning & Transfer Learning
both labelled and unlabelled data, semi-supervised data is related to the task considered.
For example, recognizing cat and dogs, transfer learning: