SEIS 763 Machine Learning - Vocabulary Flashcards (Fall 2025)

Vocabulary flashcards covering key machine learning concepts from Lecture 1 notes.

Machine Learning

Field that builds computational systems that learn from data and adapt to it; used across domains such as engineering, medicine, finance, and commerce.

Supervised Learning

ML approach that learns from labeled data to predict outputs or labels.

Unsupervised Learning

ML approach that discovers structure or patterns from unlabeled data (e.g., clustering, dimensionality reduction).

Features

Input variables used by a model to make predictions.

Label

Ground-truth target that the model aims to predict.

Model

The predictive function that maps input features to outputs.

Weights

Parameters that quantify each feature's contribution; in linear models, weights capture discriminative power.

Bias

Intercept term added to the weighted sum in linear models.

Loss Function

Function that measures how far predictions are from true labels.

Mean Squared Error (MSE)

Loss defined as the average of squared differences between predicted and true values.

Optimizer

Algorithm that updates model parameters to minimize the loss (direction of movement in parameter space).

Linear Model

Model where output is a weighted sum of inputs plus a bias: y = W^T x + b.

y_hat (Predicted)

The predicted label or value produced by the model.

y (True Label)

The ground-truth value the model tries to predict.

y = W^T x + b

Equation representing a linear regression model.

General Form yi = W^T xi + b

Linear regression equation for the i-th sample with input x_i.

Standard Data Pipeline

Common ML workflow: data collection, labeling, preparation, model building/training, evaluation, deployment.

Features Matrix X

Collection of input features organized as a matrix used by the model.

Neural Network

ML model that mimics the brain with interconnected artificial neurons, learning by adjusting connections.

Perceptron

The first artificial neural network capable of learning.

ELIZA

Early natural language processing program that simulates conversation.

The Turing Test

Test where a human evaluator interacts with a machine to determine if it can imitate a human; passing suggests AI capability.

AlphaGo

Go-playing AI developed by DeepMind that defeated a world champion.

Loss (Cost) Function

Measures error between predicted and true values; used to guide learning.