Machine Learning Glossary for Biodiversity Monitoring
This glossary provides functional definitions of machine learning terms commonly used in biodiversity monitoring. We've included Spanish translations where they exist, but some technical terms are used in both languages. These definitions aim to be practical and accessible, focusing on how terms are used in the context of ecological research and conservation, rather than providing formal technical definitions. This living document will evolve as our understanding and use of these terms develops within our community.
Este glosario proporciona definiciones funcionales de términos de aprendizaje automático comúnmente utilizados en el monitoreo de biodiversidad. Hemos incluido traducciones al español donde existen, aunque algunos términos técnicos se utilizan en ambos idiomas. Estas definiciones pretenden ser prácticas y accesibles, centrándose en cómo se utilizan los términos en el contexto de la investigación ecológica y la conservación, en lugar de proporcionar definiciones técnicas formales. Este documento vivo evolucionará a medida que nuestra comprensión y uso de estos términos se desarrolle dentro de nuestra comunidad.
| Machine Learning | Aprendizaje Automático | Technology allowing computers to learn patterns from data without explicit programming |
| Neural Network | Red Neuronal | An organization of learning functions with non-linear connections |
| Deep Learning | Aprendizaje Profundo | A machine learning technique that uses multiple neural network layers |
| Training | Entrenamiento | Process of teaching a model to make predictions using example data |
| Evaluation | evaluación | Process for measuring a models performance and skill |
| Inference | Inferencia | Process for generating predictions on new data |
| Hyperparameters | Hiperparámetros | Settings controlling the training and inference process such as learning rate, batch size |
| Learning Rate | Tasa de Aprendizaje | Controls how much the model changes in response to errors during training |
| Loss Function | Función de Pérdida | Measures how well predictions match actual data; training aims to minimize this |
| Generalization | Generalización | Model's ability to perform well on new, unseen data |
| Feature | Característica | Individual measurable property or attribute of the observed data |
| Batch Size | Tamaño de Lote | Number of training examples used in one iteration of model training |
| Epoch | Época | One complete pass through the entire training dataset |
| Computer Vision | Visión por Computadora | Technology enabling computers to interpret visual information |
| Image Recognition | Reconocimiento de Imágenes | Ability to identify objects, people, or scenes in digital images |
| Object Detection | Detección de Objetos | Technology identifying and locating specific objects within images |
| Image Classification | Clasificación de Imágenes | Process of categorizing images into predefined classes |
| Pattern Recognition | Reconocimiento de Patrones | Automated recognition of patterns and regularities in data |
| Precision | Precisión | Proportion of positive identifications that were actually correct |
| Recall | Exhaustividad | Proportion of actual positives that were correctly identified |
| Accuracy | Exactitud | Overall proportion of correct predictions |
| Processing Speed | Velocidad de Procesamiento | How quickly an AI system can analyze images or video |
| Cross-Validation | Validación Cruzada | Technique to assess model generalization to independent datasets |
| F1-Score | Puntuación F1 | Harmonic mean of precision and recall for model evaluation |
| Data Annotation | Anotación de Datos | Process of labeling data to provide ground truth for supervised learning |
| Bounding Box | Cuadro Delimitador | Rectangle drawn around an object to mark its location and size |
| Polygon Annotation | Anotación de Polígonos | Marking objects with irregular shapes using multi-point boundaries |
| Semantic Annotation | Anotación Semántica | Labeling each pixel in an image with a corresponding class |
| Instance Annotation | Anotación de Instancias | Identifying individual object instances, even within the same class |
| Keypoint Annotation | Anotación de Puntos Clave | Marking specific points of interest, often for pose estimation |
| Ground Truth | Verdad Fundamental | Accurate reference data used to train and evaluate ML models |
| Algorithm | Algoritmo | Set of rules followed by a computer to perform a task |
| Model | Modelo | Representation of what a ML system has learned from training |
| Training Data | Datos de Entrenamiento | Examples used to teach a machine learning system |
| Testing Data | Datos de Prueba | Separate examples used to evaluate model performance |
| Confidence Score | Puntuación de Confianza | Measure of how certain an AI system is about its prediction |
| False Positive | Falso Positivo | System incorrectly identifies something that isn't present |
| False Negative | Falso Negativo | System fails to identify something that is present |
| Domain Adaptation | Adaptación de Dominio | Applying a model trained on one data type to a related type |
| Data Privacy | Privacidad de Datos | Protection of personal information in ML systems |
| Bias | Sesgo | Unfair preference in AI systems, often reflecting training data bias |
| Consent | Consentimiento | Permission from individuals to use their data for AI training |
| Transparency | Transparencia | Openness about how AI systems work and make decisions |