Have you tried learning machine learning and you see a bunch of words that make you feel more lost than a dog chasing a laser pointer in a hall of mirrors?

Words like hyper-parameter, epoch, stochastic, and deterministic.

Or maybe you’re way ahead, and know what all those mean. But I know when I was first trying to learning, this was my expression:

Listening to my instructors teach ML was like nodding my head and pretending that I understood Klingon.

Pretending like I get it on the outside…crying for help on the inside.

That’s why I made this list of commonly used terminology in ML, but with simplified definitions. Maybe you know these, maybe you don’t.

But either way, these are terms you should know.

And maybe feel a bit less Neanderthal along the way. Alright, let’s go full nerd.

Algorithm: A step-by-step set of instructions a machine follows to learn from data.

Model: The end result of your algorithm’s training on the data.

Data: The raw information used to train, test, and validate your model.

Features: The pieces of your data (input variables) your model uses to make predictions. (Ex. in predicting house prices, features could be "square footage," "number of bedrooms," "location.")

Label: The correct answers or outcomes the model is trying to predict.

Training Data: The portion of your dataset used to train the model.

Testing Data: The portion of your dataset used to evaluate how well your trained model does on unseen data.

Validation Data: The separate set of data used during training to tune hyper-parameters and evaluate its performance.

Hyper-parameters: The settings you choose before training a model that influence how it learns.

Bias: An error that happens when a model makes too-simple assumptions and misses important patterns in the data.

Variance: When a model is too sensitive to changes in the training data, causing it to overfit.

Overfitting: When a model learns the training data too well, including noise, and performs poorly on new data.

Underfitting: When a model is too simple to capture the underlying patterns in the data, leading to poor performance both on training and new data.

Epoch: One complete pass through the entire training dataset.

Generalization: The ability of a model to perform well on unseen data after being trained on a specific dataset.

Stochastic: Involves randomness; outcomes can vary.

Deterministic: Predictable; same input equals same output every time.

Regression: Predicts a continuous value (e.g., house price).

Classification: Predicts a category (e.g., spam or not).

Clustering: Groups similar data without labels.

Neural Network (NN): Model with layers of connected neurons, inspired by the brain.

Deep Learning (DL): Uses neural networks with many layers.

Convolutional Neural Network (CNN): Neural network for processing images and videos.

Recurrent Neural Network (RNN): Neural network for sequential data.

Transformer: Neural network architecture for language tasks, uses attention mechanisms.

Activation Function: Adds non-linearity to neural networks (e.g., ReLU).

Backpropagation: Algorithm updating weights in neural networks.

Feature Engineering: Creating or transforming features for better models.

Dimensionality Reduction: Reduces feature number while keeping data important (e.g., PCA).

Regularization: Prevents overfitting by adding penalties (e.g., L1, L2).

Ensemble Learning: Combines models for better performance (e.g., Random Forest).

Accuracy: Ratio of correct predictions to total predictions.

Precision: Ratio of correctly predicted positives to all predicted positives.

Recall: Ratio of correctly predicted positives to all actual positives.

F1-Score: Harmonic mean of precision and recall.

Confusion Matrix: Table showing true vs. predicted values (TP, FP, TN, FN).

ROC AUC: Metric showing model's ability to classify between classes.

MAE: Average of absolute differences between predicted and actual values.

MSE: Average of squared differences between predicted and actual values.

RMSE: Square root of MSE, in the same units as target variable.

Getting fluent in machine learning lingo? Let’s put it to the test.

🧪 Interactive Google Colab: Watch the terms you just learned come to life — run a simple model on real data and see what overfitting, epochs, and training vs. testing data actually look like in practice. No prior experience needed. Just hit run and watch the machine do its thing.

📝 Practice Quiz: 3 levels. 10 questions each. Easy keeps it friendly — just the basics from today's glossary. Medium starts mixing concepts together. Hard? We're talking hyperparameters, loss functions, and the stuff that even experienced folks second-guess. No mercy.

📕 Visual Cheatsheet: A visual breakdown of every term from today's newsletter — organized by category, color-coded, and designed to be the reference you actually want to come back to. Stick it on your wall. Stare at it while you’re on the toilet. Whatever works.