More data science and AI

Binomial distribution

A distribution where each trial has only two possible outcomes.

Probability distribution

A function that describes the likelihood of obtaining the possible values that a random variable can take.

Poisson distribution

A distribution used to calculate the probability of a specific number of events occurring within a fixed interval.

Joint probability

The chance of two or more events happening together.

Marginal probability

The probability of a single event occurring, regardless of other variables.

Feature engineering

Creating, modifying, or selecting features from raw data to improve model performance.

Mean squared error (MSE)

A metric that measures the average squared difference between predicted and actual values.

Cross-validation

A method to estimate a model's performance on new data by partitioning a dataset into training and validation sets.

State-space search

Used to find solutions by exploring possible states of a problem.

Propositional logic

Simple true/false statements.

Modus Ponens

If p implies q and p is true, q must be true.

Modus Tollens

If p implies q and q is false, p must be false.

Hypothetical syllogism

Combines two implications into one (chain reasoning)

Disjunctive syllogism

If you have either/or and one is false, the other must be true.

Search algorithms

Solve problems by exploring a set of possible solutions (search space) to find an optimal path from starting state to goal state

Breadth-First-Search (BFS)

Explores graph level by level, guaranteed to find the shortest path in an unweighted graph.

Depth-First-Search (DFS)

Explores as far possible down each branch before backtracking, more memory efficient.

Heuristics

Estimate of how far a state is from the goal, guiding search to the goal efficiently. So using tactics to solve problems.

Weighted search

Trade off perfect optimality for speed.

Hierarchical Clustering

Groups data into tree-like structures of nested clusters, visualized as a dendrogram.

Apriori Algorithm

Uses breadth-first search and pruning technique to identify frequent item sets.

Policy Gradient

Trains agents by directly optimizing parameterized policy to maximize expected return in reinforcement learning.

Q-learning

Value-based, model-free reinforcement learning algorithm enabling agent to learn optimal policy for making decisions in environment.

Convolutional Neural Networks (CNN)

Deep learning model which excels at tasks involving grid-like data, such as images

Recurrent Neural Network (RNN)

Process sequential data by having a 'memory'/internal state allowing it to use past information to make current predictions.

Long Short-Term Memory (LSTM)

Advanced RNN designed to solve problems of vanishing gradients.

Transformers

Use attention to understand relationships.

Self-attention (in transformers)

Mechanism that lets each word/token decide importance of each word.

Positional Encoding

Provides sense of order, direction, and distance between words as self-attention doesn't know token order.

Personally Identifiable Information (PII)

Any information that can be linked to a specific person; can include direct identifiers like name and social security number

Interpretable surrogate models

Simpler, transparent models that mimic the behaviour of complex black-box models.

Brilliance bias

Gender stereotype, belief of higher intellectual abilities among men.

Data provenance

Historical record of a piece of data, including origin, movements, and transformations.