| Pre-trained Model | A machine learning model that has been trained on a large dataset and is ready to be fine-tuned on a specific task. Pre-trained models save time and resources as they provide a starting point that understands general features before being adapted to more specific purposes. |
| Transfer Learning | The process of improving or adapting a pre-trained model on a new, typically smaller, dataset, or task. Transfer learning leverages the knowledge a model has gained from a related task to achieve better performance or quicker convergence on a new task. |
| Fine-tuning | A specific type of transfer learning where a pre-trained model is slightly adjusted or “fine-tuned” by continuing the training process on a new dataset with potentially fewer samples or different tasks to achieve better accuracy on specific tasks. |
| Generative Adversarial Network (GAN) | A class of machine learning frameworks where two neural networks, a generator and a discriminator, are trained simultaneously to generate new data samples that are similar to a given dataset. GANs are widely used in image, video, and voice generation. |
| Transformer | A type of deep learning model that uses self-attention mechanisms to process sequences of data, such as text or time series. Transformers are the foundation of many state-of-the-art natural language processing (NLP) models. |
| BERT (Bidirectional Encoder Representations from Transformers) | A transformer-based machine learning technique for NLP tasks, including text classification, translation, and summarization. BERT’s innovation is its bidirectional training, which considers the full context of a word by looking at the words that come before and after it. |
| GPT (Generative Pre-trained Transformer) | A series of AI models designed for a variety of tasks, including but not limited to translation, summarization, and question-answering. GPT models are notable for their ability to generate coherent and contextually relevant text based on a given prompt. |
| Zero-shot Learning | A machine learning technique where a model is capable of correctly making predictions for tasks it has not explicitly been trained for, based on its understanding and generalization capabilities. |
| Few-shot Learning | A technique where a machine learning model achieves significant proficiency on a task with a very limited amount of training data, emphasizing the model’s ability to learn and adapt quickly. |
| Self-supervised Learning | A form of unsupervised learning where the data itself provides supervision. The model is trained with tasks designed so that it teaches itself the underlying structure of the data, often by predicting parts of the data from other parts. |
| Reinforcement Learning | A type of machine learning where an agent learns to make decisions by taking actions in an environment to achieve some goals. The agent learns from trial and error, receiving rewards or penalties for the actions it performs. |
| Supervised Learning | A machine learning approach where the model is trained on a labeled dataset, meaning each training example is paired with an output label. This approach is used for tasks such as classification and regression. |
| Unsupervised Learning | Machine learning techniques that learn patterns from untagged data. The system tries to learn without a teacher, identifying commonalities in the data and responding based on the presence or absence of such commonalities. |
| Semi-supervised Learning | A machine learning approach that combines a small amount of labeled data with a large amount of unlabeled data during training. Semi-supervised learning is used when acquiring a fully labeled dataset is too expensive or laborious. |
| Natural Language Processing (NLP) | The field of AI focused on enabling computers to understand, interpret, and generate human language. NLP technologies are used in a variety of applications, from chatbots and digital assistants to translation services. |
| Computer Vision | A field of AI that enables computers and systems to derive meaningful information from digital images, videos, and other visual inputs—and act on that information. |
| Neural Network | A series of algorithms that endeavors to recognize underlying relationships in a set of data through a process that mimics the way the human brain operates. Neural networks are a key technology in machine learning. |
| Deep Learning | A subset of machine learning in artificial intelligence that has networks capable of learning unsupervised from data that is unstructured or unlabeled. Deep learning is known for its ability to process large amounts of data and recognize patterns in the data. |
| Convolutional Neural Network (CNN) | A class of deep neural networks, most commonly applied to analyzing visual imagery. CNNs use a variation of multilayer perceptrons designed to require minimal preprocessing. They are also used in image and video recognition, recommender systems, and natural language processing. |
| Recurrent Neural Network (RNN) | A class of artificial neural networks where connections between nodes form a directed graph along a temporal sequence. This allows it to exhibit temporal dynamic behavior. Used in applications such as language modeling and speech recognition. |
| Attention Mechanism | A component in neural networks that weights the significance of different parts of the input data. It is crucial for models that process sequential data like text or speech, enabling the model to focus on relevant parts of the input for making decisions. |
