Python, Machine Learning, and Biomedical AI Core Concepts

Comprehensive vocabulary flashcards covering basic Python syntax, mathematical foundations for AI, machine learning and deep learning concepts, PyTorch specifics, NLP, Biomedical AI, and software/research terminology.

Variable

A named storage location in memory that holds a value or data.

Function

A reusable block of code that performs a specific action when called.

Argument / Parameter

A parameter is a variable in a function definition; an argument is the actual value passed to it.

List

An ordered, mutable (changeable) collection of items enclosed in square brackets [].

Tuple

An ordered, immutable (unchangeable) collection of items enclosed in parentheses ().

Dictionary

A collection of unordered, mutable key-value pairs enclosed in curly braces {key: value}.

Set

An unordered collection of unique elements enclosed in curly braces {}.

List comprehension

A concise, elegant way to create a new list from an existing iterable in a single line of code.

Class

A blueprint or template for creating objects, defining their attributes and behaviors.

Constructor (init)

A special method automatically called when a new object of a class is created, used to initialize attributes.

Virtual environment

An isolated workspace that allows you to install specific versions of packages for a project without affecting other projects.

Scalar

A single number representing magnitude (e.g., 5 or -1.2).

Vector

An ordered 1D array of numbers representing both magnitude and direction.

Matrix

A 2D grid or array of numbers organized into rows and columns.

Tensor

A multi-dimensional array of numbers (0D is a scalar, 1D is a vector, 2D is a matrix, 3D+ is a tensor).

Shape

A tuple showing the number of elements along each dimension of a tensor (e.g., a $3 \times 4$ matrix).

Transpose

An operation that flips a matrix over its diagonal, swapping its rows and columns.

Identity matrix

A square matrix with ones on the main diagonal and zeros everywhere else, acting like the number 1 in matrix algebra.

Derivative

The rate of change of a function with respect to a variable; geometrically, the slope of a curve.

Gradient

A vector of partial derivatives pointing in the direction of the steepest ascent of a multi-variable function.

Chain rule

A formula for calculating the derivative of a composite function (a function inside another function).

Global minimum

The absolute lowest possible point across the entire domain of a function.

Mean

The average value of a dataset, calculated by dividing the sum of all values by the total count.

Variance

A measure of how far a set of numbers is spread out from their mean value.

Normal distribution

A symmetric, bell-shaped probability distribution where most observations cluster around the central mean.

Overfitting

A flaw where a model learns training data too well (including its noise), causing poor performance on new, unseen data.

Feature

An individual measurable property or input variable used to make a prediction (e.g., tumor size).

Label

The ground-truth answer or outcome associated with a data sample.

Training set

The portion of the dataset used to teach the model and adjust its internal parameters.

Validation set

A separate portion of data used to tune hyperparameters and evaluate the model during training to prevent overfitting.

Classification

An ML task that predicts a discrete, categorical class or label (e.g., "Benign" or "Malignant").

Regression

An ML task that predicts a continuous, numerical value (e.g., life expectancy in months).

Epoch

One complete pass of the entire training dataset through the machine learning model.

Learning rate

A small multiplier hyperparameter that controls how drastically a model adjusts its weights during optimization.

Loss function

A mathematical formula that measures the error of a model's prediction for a single data sample.

Backpropagation

An algorithm that calculates the gradients of the loss function with respect to the model's weights, working backward from output to input.

Precision

The ratio of true positive predictions to the total number of predicted positives (measures reliability of positive calls).

Recall

The ratio of true positive predictions to the total number of actual positives (measures ability to find all positive cases).

F1-score

The harmonic mean of precision and recall, providing a balanced metric for imbalanced datasets.

Confusion matrix

A tabular layout showing the breakdown of True Positives, True Negatives, False Positives, and False Negatives.

Neuron

The fundamental computational unit of a neural network that receives inputs, applies weights, adds a bias, and passes the result through an activation function.

Activation function

A non-linear mathematical function applied to a neuron's output to enable the network to learn complex patterns.

ReLU (Rectified Linear Unit)

An activation function that outputs the input directly if it is positive, and outputs zero if it is negative ($f(x) = \max(0, x)$).

Softmax

An activation function applied to the final layer that converts a vector of raw scores into a probability distribution that sums to 1.

CNN (Convolutional Neural Network)

A type of network specialized for processing grid-like data like images, utilizing filters to extract spatial features.

RNN (Recurrent Neural Network)

A type of network designed for sequential data (like text or time series) that maintains internal memory loops.

Transformer

A state-of-the-art architecture based on self-attention mechanisms, widely used for processing entire sequences of data in parallel (e.g., LLMs).

Dropout

A regularization technique where random neurons are ignored/disabled during training to prevent co-dependency.

Broadcasting

PyTorch's ability to automatically expand smaller tensors to match the shape of larger tensors during arithmetic operations.

nn.Linear

A built-in PyTorch layer that applies a standard fully connected linear transformation ($y = xW^T + b$).

Adam

An advanced optimizer that computes adaptive learning rates for each individual parameter based on historical updates.

CUDA

NVIDIA’s parallel computing platform and API that allows PyTorch to utilize GPUs for massive acceleration.

State dict

A standard Python dictionary mapping each layer of a PyTorch model to its corresponding parameter tensors (weights/biases).

Tokenization

The process of splitting raw text strings into discrete tokens and mapping them to numerical IDs.

Embedding

A dense vector representation of a token that captures its semantic meaning and relationship to other tokens.

Self-attention

An attention mechanism where tokens in a sequence calculate their contextual relationships with all other tokens in the same sequence.

BERT

An encoder-only Transformer model built by Google, optimized for understanding the contextual meaning of words bidirectionally.

Biomarker

A biological molecule found in blood, fluids, or tissues that is a measurable sign of a normal process, condition, or disease.

FASTQ

A standard text-based file format used to store nucleotide sequences from DNA/RNA sequencing instruments along with quality scores.

Kaplan-Meier curve

A non-parametric statistic/graph used to estimate and visualize the survival probability of patients over time.

API (Application Programming Interface)

A software intermediary that allows two separate applications or systems to communicate and exchange data with each other.

Docker

A tool designed to easily create, deploy, and run applications by using isolated software packages called containers.

Ablation study

A research experiment where parts of an AI architecture are systematically removed to understand the isolated contribution of each component.

p-value

The probability value used in hypothesis testing; a $p\text{-value} < 0.05$ typically indicates that results are statistically significant.