Accuracy
The ratio of correctly predicted observations to the total observations.
Activation Function
Determines the output of a neural node, introducing non-linearity to the model.
Adversarial Neural Network
A network trained to generate data by competing against another network.
Algorithm
A set of rules or instructions given to an AI, ML, or DS system to help it learn on its own.
Anomaly Detection
Identifying patterns in data that do not conform to expected behavior.
Artificial Neural Network (ANN)
Computing systems inspired by the human brain’s structure, fundamental to deep learning.
Attention Mechanisms
Techniques in deep learning that allow models to focus on specific parts of the input, enhancing the model’s ability to remember long-term dependencies.
Autoencoder
A neural network used for unsupervised learning of efficient codings, often for dimensionality reduction or anomaly detection.
Backpropagation
An algorithm used for training neural networks, based on minimizing the error in the outputs.
Bagging
An ensemble method that averages the predictions of multiple models to improve accuracy and reduce overfitting.
Batch Learning
Training a model with the entire dataset at once, rather than incrementally.
Bayesian Networks
A probabilistic graphical model that represents a set of variables and their conditional dependencies.
BERT (Bidirectional Encoder Representations from Transformers)
A transformer-based machine learning technique for NLP pre-training.
Bias
An error introduced in your model due to oversimplification of the machine learning algorithm.
Bias (in data)
Systematic error introduced by a sampling or data collection process.
Boosting
An ensemble technique that adjusts the weight of an observation based on the last classification.
Capsule Networks
Neural networks that aim to recognize patterns in an image without losing spatial hierarchies.
Capsule Networks (CapsNets)
A type of deep learning system that aims to improve the shortcomings of CNNs, especially in recognizing spatial hierarchies between features.
Computer Vision
A field of AI that trains machines to interpret and make decisions based on visual data.
Confusion Matrix
A table used to evaluate the performance of a classification model, showing actual vs. predicted classifications.
Contrastive Learning
A self-supervised learning technique where the model is trained to distinguish between similar and dissimilar data points.
Classification
A supervised learning approach where the output variable is a category, like “spam” or “not spam”.
Clustering
An unsupervised learning approach that groups sets of data with similar patterns.
Cross-Validation
A technique to assess how a model will generalize to an independent dataset by partitioning the original sample.
DALL·E
A variant of the GPT-3 model by OpenAI, designed to generate images from textual descriptions.
Data Augmentation
Techniques to artificially increase the size of a training dataset by modifying the data.
Data Imputation
The process of replacing missing data with substituted values.
Data Mining
The process of discovering patterns and knowledge from vast amounts of data.
Data Normalization
Adjusting data into a standard scale, often between 0 and 1.
Decision Tree
A flowchart-like structure used for decision-making, where each node represents a feature.
Deep Learning
A subset of ML that uses neural networks with many layers (hence “deep”).
Dimensionality Reduction
The process of reducing the number of random variables under consideration by obtaining a set of principal variables.
Dropout
A regularization technique for neural networks where randomly selected neurons are ignored during training.
Early Stopping
A technique to stop training once the model performance stops improving on a held-out validation dataset.
EDA (Exploratory Data Analysis)
An initial step in data analysis, used to summarize main characteristics with visual methods.
Ensemble
Combining multiple machine learning models to improve accuracy and robustness.
Ensemble Learning
Combines multiple models to produce a single predictive model for improved accuracy.
Ensemble Methods
Techniques that create multiple models and then combine them to produce improved results.
Embedding Layer
Used in deep learning, it converts sparse categorical data into a dense vector representation.
Epoch
One complete forward and backward pass of all the training examples in neural network training.
Evolutionary Algorithm
Algorithms inspired by the process of natural selection, used for optimization.
Explainable AI (XAI)
Techniques and methods to create a more transparent and interpretable AI system.
F1 Score
The weighted average of Precision and Recall, providing a balance between the two.
Feature
An individual measurable attribute or property for a phenomenon being observed.
Feature Engineering
The process of selecting, transforming, or creating relevant input variables to improve model performance.
Feature Scaling
Techniques like normalization and standardization to make sure features have similar scales.
Federated Learning
A machine learning structure where the model is trained across multiple devices or servers while keeping the data localized.
Few-shot Learning
Training a model to learn with a very small labeled dataset.
Fine-tuning
Making slight adjustments to a model to improve its performance.
Genetic Algorithm
A search heuristic inspired by the process of natural selection, used to find approximate solutions to optimization and search problems.
Generative Adversarial Network (GAN)
Consists of two networks, one generating data and the other evaluating it.
GPT (Generative Pre-trained Transformer)
A series of language representation models. ChatGPT, for instance, is known for its large scale and versatility in various tasks without task-specific training.
Gradient Descent
An optimization algorithm used to minimize the function that represents the model’s error.
Graph Neural Networks (GNN)
Neural networks designed to process data structured as graphs.
Grid Search
An exhaustive search over a specified parameter space to find the best model hyperparameters.
Heuristic
A problem-solving approach that uses shortcuts for producing good-enough solutions in a reasonable time.
Hidden Layer
Layers between the input and output layers in a neural network where feature learning occurs.
Hyperparameter
Parameters in the model that are set before training (as opposed to during training).
Instance-Based Learning
A type of learning where the model learns from instances of the training data rather than a general model (e.g., k-Nearest Neighbors).
Inference
The process of making predictions using a trained model.
Input Layer
The layer of the neural network that receives input from the dataset.
Jupyter Notebook
An open-source tool that helps you create and share documents that contain live code, equations, and visualizations.
K-means
An unsupervised clustering algorithm that divides a dataset into ‘k’ number of clusters.
Keras
An open-source software library that provides a Python interface for neural networks.
Knowledge Distillation
The process where a compact model is trained to replicate the behavior of a larger model (or an ensemble), aiming to transfer knowledge from the larger model to the smaller one.
Label
The “answer” or “result” for a data point in supervised learning.
LangChain
An Orchestration framework designed to simplify the creation of applications using large language models (LLMs).
Large Language Model
A machine learning model trained on vast amounts of text data to generate human-like text.
Learning Rate
A hyperparameter that determines the step size at each iteration while moving towards a minimum of the loss function.
Linear Regression
A statistical method to model the relationship between a dependent variable and one or more independent variables.
Long Short-Term Memory (LSTM)
A type of RNN designed to recognize patterns over long intervals of time.
Logistic Regression
Used for binary classification, it estimates the probability that a given instance belongs to a particular category.
Loss Function
A measure of how well a model’s predictions match the true values; it’s what algorithms aim to minimize.
Machine Learning
A method of data analysis that automates analytical model building.
Markov Decision Process (MDP)
A mathematical framework for modeling decision-making in situations where outcomes are partly random and partly under the control of a decision-maker.
MLOps
Practices for collaboration and communication between data scientists and operations professionals to automate the end-to-end machine learning lifecycle.
Model
In machine learning, it represents what was learned from a dataset.
Multi-Layer Perceptron (MLP)
A type of feedforward neural network with multiple layers between input and output.
Naive Bayes
A classification technique based on Bayes’ theorem with an assumption of independence among predictors.
Neuron
A single node in a neural network.
Neural Architecture Search (NAS)
An automated process for finding the best neural network architecture.
Neural Radiance Fields (NeRF)
A method for synthesizing novel views of complex scenes by optimizing a simple fully connected neural network.
Normalization
The process of scaling input data to a standard scale.
Optimization Algorithms
Algorithms used to adjust model parameters to minimize the model error.
One-shot Learning
Training a model to recognize patterns with very limited data.
Overfitting
When a model is too closely adapted to the training set and performs poorly on new, unseen data.
Outlier
A data point that differs significantly from other observations and can affect the results of an analysis.
Output Layer
The layer that produces the result for given inputs.
Parameter
Internal configurations of a model that are learned from the data during training (e.g., weights in a neural network).
Perceptron
A type of artificial neuron which is the basic operational unit of an ANN.
Precision
The ratio of correctly predicted positive observations to the total predicted positives.
Pooling Layer
Used in CNNs, it reduces the spatial dimensions of the data while retaining important features.
Principal Component Analysis (PCA)
A method used to emphasize variation and bring out strong patterns in a dataset.
Q-Learning
A type of model-free reinforcement learning algorithm.
Quantum Computing
A type of computation that uses qubits to represent data, offering potential advantages for specific types of computations.
Quantum Neural Networks (QNN)
Neural networks that are designed to run on quantum computers.
Query
A request for data retrieval from a database.
Random Forest
An ensemble of decision trees, often trained with the “bagging” method.
Recall (Sensitivity)
The ratio of correctly predicted positive observations to all the actual positives.
Regularization
Techniques used to prevent overfitting by adding a penalty to the loss function.
Reinforcement Learning
A type of machine learning where agents learn how to behave by receiving rewards for good actions.
ReLU (Rectified Linear Unit)
A type of activation function that is defined as the positive part of its argument.
Responsible AI
The practice of designing, building, and deploying AI in a manner that is transparent, accountable, and beneficial for all.
Self-Supervised Learning
A training paradigm where the model generates its own supervisory signal from the input data.
Semi-Supervised Learning
Uses both labeled and unlabeled data for training, typically a small amount of labeled and a large amount of unlabeled data.
Sequence-to-Sequence Model
Used in NLP tasks like translation, it maps input sequences to output sequences.
Softmax Function
A function that takes an unnormalized vector and normalizes it into a probability distribution.
Supervised Learning
A type of machine learning where the algorithm is trained on labeled data.
Support Vector Machine (SVM)
A supervised learning model used for classification and regression analysis.
Tensor
A mathematical object analogous to vectors and matrices and is used in deep learning frameworks like TensorFlow.
Tokenization
The process of converting text into individual tokens or words.
Transfer Learning
A technique where a pre-trained model is fine-tuned for a different but related task.
Transformers
A deep learning model architecture, particularly in NLP, that uses attention mechanisms to capture contextual information from input data.
Transformer Architecture
A deep learning model architecture, primarily used in NLP, known for its self-attention mechanism.
Turing Test
A measure of a machine’s ability to exhibit intelligent behavior indistinguishable from that of a human.
Unsupervised Learning
Machine learning where the model is not provided with the correct results during training.
Underfitting
When a model is too simple to capture the underlying structure of the data.
Validation Set
A subset of the data used to evaluate a model’s performance during training.
Vector
A quantity having direction and magnitude, often used in machine learning models.
Weights
The parameters in neural networks that transform input data within the network’s layers.
Weight Initialization
The practice of setting the initial values of the weights in a neural network before training.
Word Embedding
Represents words as vectors in a high-dimensional space, capturing semantic relationships.
XGBoost
An optimized distributed gradient boosting library.
X-axis
In data visualization, it’s often the horizontal axis used to plot data points.
XGBoost
An optimized gradient boosting library designed to be highly efficient, flexible, and portable.
Y-axis
In data visualization, it’s often the vertical axis used to plot data points.
YARN (Yet Another Resource Negotiator)
A resource-management platform responsible for managing computing resources in clusters.
Zero-shot Learning
Training a model to handle tasks it hasn’t seen during training.
Z-score
A statistical measurement that describes a value’s relationship to the mean of a group of values.