![[Pasted image 20240611142732.png|525]]
# Machine Learning
The “learning” in **machine learning** refers to trying to find hidden patterns within a set of data. There are
In supervised learning, the data
The actual concept of machine learning is extremely broad
In unsupervised learning
- *Exploratory data analysis:* We may want to explore the data without a particular end goal in mind.
---
> [!abstract]+ **Common Terminology**
> **Overfitting** is when a created model sticks too closely to the original data set and significantly lowers its accuracy when applied to data outside of the training set.
>
> **Hyperparameters** are external configuration variables that data scientists use
>
## Types of Machine Learning
| Type | Description | Example |
| ------------------- | ----------- | ------------------------------------------ |
| Classification | | Fraud detection & covid classification |
| Clustering | | Customer segmentation & targeted marketing |
| Dimension Reduction | | Single cell |
| Regression | | |
- Machine Learning
- Unsupervised
- Clustering
- customer segmentation
- targeted marketing
- Dimensionality Reduction
- structure discovery
- meaningful compression
- Supervised
- Regression
- sales forcasting
- weather forcasting
- Classification
- fraud detection
- covid classification
- Reinforcement
- Robot Navigation
- Gaming AI
- Random Forest (mean of multiple decision trees)
- Support Vector Machines (SVM): plotting a linear line across potentially high dimensions of data
- Canonical Correlation Analysis (CCA):
**Deep Learning**
- Most notable example are *neural networks*
- Gets its name from the many different layers within the model that imitate the networks of neurons found within brain tissue
-
## Supervised Learning
- Given a data set of pairs
- *Creates model* to predict future outcome from input data
- Examples:
- Linear Regression (square error)
### Types of Data Labels
Labels can be
### Reinforcement Learning
**Reinforcement Learning**
- Based on rewarding desired behaviors and punishing undesired ones
- [ ] vs adversarial neural networks?
## Unsupervised Learning
- Given a data set and expected to discover notable patterns within it
- Examples:
- [[Clustering Algorithms|k-means clustering]]
- [[Dimension Reduction|principal component analysis (PCA)]]
A large majority (but not all) of unsupervised learning algorithms are non-deterministic
## Properties of Models
### Deterministic vs Probabilistic
> See also:
> - [[Probability Theory]]
A machine learning model is **deterministic** if
On the other hand, a model is called **probabilistic** if it
- One example of a highly probabilistic model
## Components of Machine Learning Models
### Processing Input Data
#### Transformers
### Domains
- Source Domain:
- Target Domain: