Training Methods58

How AI learns from data. Supervised, unsupervised and reinforcement learning, fine-tuning, overfitting, loss functions, gradient descent: the steps hidden behind the word "training", one at a time.

Level Beginner 13 Intermediate 31 Advanced 14

Agent & Environment

The side that decides, and the stage that reacts to it

Beginner Alignment

Matching what an AI is capable of to what people actually want from it

Ethics Intermediate Backpropagation

Tracing error backward to split blame across the network

Intermediate Catastrophic Forgetting

Losing old skill while learning something new

Advanced Checkpoint

A saved snapshot of training at one point in time

Beginner Clustering

Grouping similar things together with no labels at all

Basics Intermediate Contrastive Learning

Learning by pulling alike things together and pushing others apart

Advanced Convergence

When training stops improving and settles in place

Intermediate Cross-Validation

Splitting data into pieces, checking each in turn, then averaging

Advanced Data Augmentation

Transforming data you already have to multiply how much of it you have

Intermediate Data Leakage

When information that should stay separate crosses into training

Evaluation Intermediate DQNDeep Q-Network

A neural network standing in for a table of expected scores

Advanced Dropout

Randomly resting part of a model during training

Intermediate Early Stopping

Stopping training right when the score starts to slip

One full pass through all the training data

Beginner Exploration vs Exploitation

The balance between using what you know and trying something new

Intermediate Federated Learning

Skipping the data pool, gathering only what each side learned

Advanced Fine-Tuning

Nudging a trained model further with new data

Intermediate Generalization

The ability to keep performing well on data never seen before

Basics Intermediate Genetic Algorithm

Mixing and nudging a batch of candidates to pick the best

Intermediate Gradient

The signal that says which way, and how far, to reduce error

Architecture Intermediate Gradient Descent

Nudging values downhill toward less error, one small step at a time

Intermediate Hyperparameter

A setting a person picks before training begins

Intermediate Imitation Learning

Watching someone skilled and copying their moves exactly

Intermediate Instruction Tuning

Teaching a model to answer in the shape a request calls for

LLM Intermediate Knowledge Distillation

Pouring a big model's judgment into a smaller one

Advanced Learning Curve

A line showing how the score changes over training

Beginner Learning Rate

How far a single training step moves the values

Intermediate Local Minimum

A spot that's stuck low, even though lower ground exists

Intermediate LoRA

Leaving the main body alone and training only a small add-on

Advanced Loss Function

A single number for how far off the answer was

Intermediate Machine Learning

Finding rules from examples instead of writing them by hand

Basics Beginner Mean Squared Error

A penalty that squares how far off you were, then averages

Evaluation Intermediate Mini-Batch

Splitting data into chunks and processing one chunk per step

Intermediate Mode Collapse

When the generator settles on one winning output and just repeats it

Generative Advanced Model Collapse

When AI keeps learning from its own output, variety quietly shrinks

Ethics Advanced Noise

The blur mixed in — and also where generation starts

Generative Beginner Optimizer

The style of moving values once the gradient points the way

Advanced Overfitting

Memorizing the practice answers and failing on anything new

Intermediate Policy

A standing rule for what to do in each situation

Intermediate Pretraining

Learning the basics before anyone decides what the job will be

LLM Intermediate Q-Learning

Writing down an expected score for every action and refining it

Advanced Regression

Naming a single number somewhere on a continuous scale

Basics Intermediate Regularization

A device that fines a model whenever its values grow too large

Intermediate Reinforcement Learning

Learning by trying things and adjusting to the score you get back

Beginner Reward

The single score that comes back after an action, good or bad

Polishing the texture of answers using what people prefer

LLM Intermediate Scaling Law

The rule for how much better performance gets as scale grows

LLM Intermediate Self-Supervised Learning

Learning by having the data make its own problems and answers

Intermediate Semi-Supervised Learning

Learning from a little answered data and a lot that isn't

Advanced Supervised Learning

Learning a rule from examples that already carry the right answer

Beginner Training

The process of nudging dials a little at a time from examples

Basics Beginner Transfer Learning

Carrying a skill already learned over to a new task

Intermediate Underfitting

Too simple to get even the practice data right

Intermediate Unsupervised Learning

Finding structure in data with no answers attached

Beginner Validation Set

Data set aside for checking only, never used to train

Beginner Vanishing Gradient

When the fix-it signal fades away layer by layer, moving backward

Architecture Advanced World Model

A model of how the world works that you can run before acting

Generative Advanced