Lecture 18 - AI & ML Flashcards
define: artificial intelligence
any computational task that requires intelligence comparable to or greater than a human’s
define: machine learning
use of statistical algorithms that transform input data into generalized models, which can then be used to perform tasks without being explicitly given instructions
what are the types of machine learning
- supervised learning
- unsupervised learning
- self-supervised learning
- reinforcement learning
define: supervised learning
training a model using labelled data: a set of paired inputs and outputs which are known beforehand
what are common use cases of supervised learning
- classification: predicts a discrete value
- regression: predicts a continuous numerical value
what are two major types of SL architecture
- decision trees
- neural networks
what are decision trees
a simple type of ML algorithm that arranges a set of known input variables, or features, in a hierarchical tree structure
when are decision trees not ML
if its not made based on data
what are random forests (RF)
an ensemble learning method (combines multiple sub methods) that generates multiple decision trees
what is the output of random forests
- classification: the most common prediction
- regression: the average of all predictions
what is the advantage of random forests over a single decision tree
higher accuracy, less bias
what are RFs a popular tool for
genetic association studies
what does deep learning use
multi-layered neural networks
how are neural networks evaluated
layer-by-layer
what are the values of the nodes in a layer of a neural network
the sum of contributions from all nodes from the previous layer
define: unsupervised learning
analyzing unlabeled data and finding patterns and relationships within it
when is unsupervised learning used
- clustering
- dimensionality reduction
what is the architecture of unsupervised learning
- clustering algorithms
- some types of neural networks
define: self-supervised learning
trained on unlabeled data and it creates its own ‘labels’ based on features within the training data, and uses them to make predictions
what is Natural Language Processing a form of
self-supervised learning
what is ‘-omics’ data analysis a form of
self-supervised learning
define: reinforcement learning
training by repeatedly interacting with an environment, receiving feedback (in the form of rewards/penalties)
what is the goal of reinforcement learning
learning how to make decisions that maximize the reward signal
what are the use cases of reinforcement learning
- exploration: trying new options
- exploitation: choosing the best option based on current knowledge
