Learning a mapping from labelled $(x, y)$ pairs. ML Lec 2

An OLS model predicting a continuous target as a linear function of the features. DS Lec 4

Ordinary Least Squares — the estimator that minimises the sum of squared residuals. DS Lec 4

A scalar measure of model error (e.g. MSE) that training minimises. ML Lec 2

Iterative first-order optimisation that steps the parameters downhill on the Cost Function. ML Lec 2

The closed-form OLS solution $(\mathbf{X}^\top\mathbf{X})^{-1}\mathbf{X}^\top\mathbf{y}$. ML Lec 2

A linear model fit on engineered polynomial features. ML Lec 2

Putting features on a comparable scale (min-max scaling, standardisation). ML Lec 2

Penalising model complexity — L1 (Lasso), L2 (Ridge), Elastic Net, early stopping. ML Lec 2

L2-regularised linear regression. DS Lec 4

L1-regularised regression that performs automatic Feature Selection. DS Lec 4

A large generalisation gap — the model fits noise; the heart of the bias-variance trade-off. ML Lec 2

The trade-off between underfitting (high bias) and overfitting (high variance). DS Lec 4

Root Mean Squared Error — a regression error metric in the target's units. DS Lec 4

The coefficient of determination — the proportion of target variance the model explains. DS Lec 4

A supervised task mapping inputs to discrete labels (binary, multiclass, multilabel). ML Lec 5

K-Nearest Neighbours — a lazy classifier using a majority vote among the $k$ closest training points. ML Lec 5

A flowchart classifier that splits on features via if/else rules. DS Lec 3

An ensemble of decision trees with bagging plus feature randomness. DS Lec 3

A probabilistic binary classifier using the Sigmoid Function. DS Lec 3

Maps real values to $[0,1]$, giving logistic regression its probability output. DS Lec 3

A node-purity measure used to choose decision-tree splits. DS Lec 3

Combining multiple models to lower variance (e.g. bagging in a Random Forest). DS Lec 3

The similarity measure KNN relies on (Euclidean, Manhattan, Minkowski). ML Lec 5

The standard straight-line distance used by K-Means and KNN. DS Lec 5

Encoding a categorical variable as a set of binary columns. DS Lec 3

Choosing the model type and tuning hyperparameters for a given task. ML Lec 5

Partitioning data into a training set and a held-out test set for honest evaluation. ML Lec 5

A third partition used to compare models and tune hyperparameters without touching the test set. ML Lec 5

Selecting hyperparameter values (e.g. $k$) via a validation-set search. ML Lec 5

K-fold resampling for unbiased model comparison and hyperparameter tuning. ML Lec 7

The fraction of correct predictions on the test set. ML Lec 5

A table of TP, TN, FP (Type I) and FN (Type II) classification outcomes. ML Lec 7

$TP / (TP + FP)$ — minimises false positives. ML Lec 7

$TP / (TP + FN)$ — minimises false negatives; also called TPR / sensitivity. ML Lec 7

The paired evaluation metrics for imbalanced classification — see Precision and Recall. DS Lec 3

TPR vs FPR across all thresholds — measures classifier separability. ML Lec 7

Area under the ROC curve — the probability the model ranks a random positive above a random negative. ML Lec 7

Using test-set information during training, which invalidates the evaluation. ML Lec 7

Chaining preprocessing and model into one object to prevent Data Leakage. DS Lec 4

Finding structure in unlabelled data. DS Lec 5

A centroid-based clustering algorithm. DS Lec 5

Plotting within-cluster sum of squares vs $K$ to pick the number of clusters. DS Lec 5

Measures how well each point fits its assigned cluster. DS Lec 5

Principal Component Analysis — unsupervised feature extraction via a variance-maximising rotation. ML Lec 7

The proportion of total variance captured by the PCA components. DS Lec 5

Exponential data sparsity (and degrading distances) in high-dimensional spaces, motivating dimensionality reduction. ML Lec 7

A business application of clustering to group customers. DS Lec 5

Choosing a subset of the original features (wrappers and filters). ML Lec 7

A linear-dependency measure $\rho = \sigma_{xy}/(\sigma_x \sigma_y)$, used as a filter criterion. ML Lec 7

The workflow of summarising, visualising, and sanity-checking a dataset before modelling. DS Lec 2

The Python library for tabular data manipulation — the DataFrame toolkit. DS Lec 2

Pandas' 2-D labelled table (rows × named columns), the core EDA data structure. DS Lec 2

The split-apply-combine pattern: split rows by a key, apply an aggregation, combine the results. DS Lec 2

Measures of central tendency (mean, median, mode) and spread (standard deviation, quartiles) describing a column. DS Lec 2

The shape of a variable's values (symmetric, skewed), read from a histogram. DS Lec 2

A chart that bins a numeric variable to show its distribution. DS Lec 2

A chart showing the median, quartiles, and outliers via the IQR. DS Lec 2

Absent entries (NaN) that must be detected and handled (drop or impute) before modelling. DS Lec 2

Extreme values — flagged by the IQR rule — that can distort summaries and models. DS Lec 2

The foundational Python plotting library. DS Lec 2

A higher-level statistical plotting library built on Matplotlib. DS Lec 2

The caution that a statistical association does not by itself imply a causal effect. DS Lec 2