AI & Machine Learning Concepts Study Guide: 97 Flashcards From Backpropagation to Transformers

Machine learning has a vocabulary problem: a single lecture can throw backpropagation, dropout, F1 score and Markov decision process at you in ten minutes, and each one hides a precise definition. The AI & Machine Learning Concepts deck on Flip It (deck code XS5846, 97 cards) turns that vocabulary into term → definition flashcards, written at the level of an intro ML course or a technical interview screen.

Here is what is inside and how to study it so the definitions become usable, not just recognisable.

A sample of the 97 cards

Concepts covered, by theme

• Learning paradigms: AI, machine learning, deep learning, supervised / unsupervised / reinforcement learning, classification vs regression, transfer learning, fine-tuning, few-shot and zero-shot learning.
• Neural network mechanics: artificial neuron, activation function (ReLU, sigmoid, tanh), loss function, gradient descent and SGD, learning rate, batch size, epoch, hyperparameters, early stopping, data augmentation, GPU acceleration, TensorFlow, PyTorch.
• Architectures: CNN, RNN, LSTM, transformer, attention mechanism, GAN, VAE, diffusion models, BERT, GPT, large language models, prompt engineering.
• Generalization and evaluation: overfitting, underfitting, L1/L2 regularization, dropout, cross-validation and k-fold, precision, recall, F1, accuracy, confusion matrix, ROC curve, AUC, bias, variance, bias-variance tradeoff.
• Classical algorithms: linear and logistic regression, decision tree, random forest, gradient boosting (XGBoost, LightGBM), ensemble learning, SVM, k-nearest neighbors, naive Bayes, k-means, clustering, PCA, t-SNE.
• Data preparation: feature engineering, feature scaling, one-hot encoding, dimensionality reduction, tokenization, word embeddings.
• Application fields: NLP, sentiment analysis, named entity recognition, machine translation, computer vision, image classification, object detection (YOLO, R-CNN), image segmentation, facial recognition, generative AI.
• Reinforcement learning: agent, policy, exploration vs exploitation, Markov decision process, Q-learning.
• Big picture: AGI, the Turing test, explainable AI, AI ethics, AI winters, the current AI boom.

Study tips for this deck

Learn the metrics as one confusion matrix, not five formulas. Precision, recall, accuracy, F1, ROC and AUC are six separate cards, but they all derive from the same four cells (TP, TN, FP, FN). Draw the 2×2 grid once, write each formula next to it, and the cards stop feeling like independent facts: precision reads down the predicted positive column, recall across the actual positive row, accuracy is the diagonal over everything.

Group the "opposite pairs" and review them back to back. This deck is full of contrasts: overfitting / underfitting, bias / variance, supervised / unsupervised, classification / regression, exploration / exploitation, few-shot / zero-shot, CNN (grids) / RNN (sequences), GAN (generator vs discriminator) / VAE (compressed representation). Studying each pair together forces you to articulate the difference, which is exactly what interviewers ask.

Reverse the cards for the acronyms. For LSTM, PCA, SVM, KNN, NER, GAN, VAE, AUC and AGI, quiz yourself from the definition back to the abbreviation — the expansion is trivial, the recognition is what fails under pressure.

Get the deck

Log in at flip-it.app and search for XS5846 or "AI & Machine Learning Concepts". Spaced repetition will bring back learning rate and gradient boosting just before you forget them. If you are building a data toolkit alongside the theory, the Pandas Essentials study guide covers the DataFrame skills you will need to actually feed these models.