New to Gen AI? Read our Generative AI guide.
Generative AI Glossary
- Autoencoder: A type of neural network that learns a compressed representation of data, which can be used for generation and other tasks.
- Backpropagation: A process used in neural networks to adjust the weights of the network based on the error between predicted and actual outputs.
- BERT: Bidirectional Encoder Representations from Transformers - a type of LLM developed by Google that is pre-trained on large amounts of text data.
- BigGAN: A type of GAN developed by Google that is trained on large amounts of data to generate high-quality images.
- Capsule Network: A type of neural network architecture that uses capsules to represent entities in data, and is able to handle variations in viewpoint and pose.
- DeepDream: A technique developed by Google that uses neural networks to generate hallucinogenic images from existing images.
- Dropout: A technique used in neural networks to prevent overfitting by randomly dropping out certain nodes during training.
- Encoder: A component of a neural network that transforms input data into a compressed representation.
- Fine-tuning: A process in which a pre-trained generative model is further trained on a specific task or data set.
- Fully Connected Layer: A type of neural network layer in which every input is connected to every output.
- Generative Design: A design process in which a generative model is used to generate and evaluate potential designs.
- Gradient Descent: A process used in neural networks to minimize the loss function by adjusting the weights of the network in the direction of the negative gradient.
- Hyperparameter: A parameter of a machine learning model that is set before training and determines the model's structure and behavior.
- Inception Model: A deep neural network developed by Google that is used for image classification and other tasks.
- Interpolation: A process in which a generative model is used to generate new data that lies between existing data points.
- Latent Space: The compressed representation of data learned by a generative model.
- Leaky ReLU: A variant of the rectified linear unit activation function that allows a small amount of negative input to pass through.
- Logistic Regression: A type of machine learning algorithm used for binary classification.
- Loss Function: A function used to measure the difference between predicted and actual outputs in a machine learning model.
- Multi-Layer Perceptron: A type of neural network architecture consisting of multiple layers of fully connected nodes.
- One-Shot Learning: A type of machine learning in which a model is trained to recognize new objects from only a few examples.
- Outlier: A data point that is significantly different from other data points in a set.
- PixelCNN: A generative model that is able to generate images pixel by pixel.
- Recurrent Neural Network: A type of neural network architecture that is able to handle sequential data, such as text or time series data.
- ResNet: A type of deep neural network architecture that uses residual connections to improve training and performance.
- Reverse Image Search: A technique that uses a generative model to search for similar images to a given input image.
- Self-Attention: A mechanism used in neural networks to allow the network to focus on different parts of the input data.
- Seq2Seq: A type of neural network architecture used for sequence-to-sequence tasks, such as machine translation.
- Style Transfer: A technique that uses a generative model to transfer the style of one image onto another.
- Supervised Learning: A type of machine learning in which the model is trained on labeled data, with the goal of predicting labels for new data.
- Tensor: A mathematical object used in machine learning to represent multi-dimensional arrays of data.
- Transformer: A type of neural network architecture that uses self-attention to allow the network to focus on different parts of the input data.
- Transfer Learning: A process in which a pre-trained machine learning model is adapted to a new task or data set.
- Unsupervised Learning: A type of machine learning in which the model is trained on unlabeled data, with the goal of discovering patterns and structures in the data.
- Variational Inference: A method used to approximate complex probability distributions, often used in generative models.
- Weight Decay: A technique used in neural networks to prevent overfitting by penalizing large weights.
- Word Embedding: A technique used to represent words as low-dimensional vectors, often used in natural language processing tasks.
- Zero-Shot Learning: A type of machine learning in which a model is able to recognize new objects without any examples.
- Zombie AI: A hypothetical scenario in which an AI system becomes uncontrollable or poses a threat to humanity.