Prioritised Experience Replay is a technique used in machine learning, particularly in reinforcement learning, to improve how an algorithm learns from past experiences. Instead of treating all previous experiences as equally important, this method ranks them based on how much they can help the learning process. The algorithm then focuses more on experiences that are…
Category: Deep Learning
Dueling DQN
Dueling DQN is a type of deep reinforcement learning algorithm that improves upon traditional Deep Q-Networks by separating the estimation of the value of a state from the advantages of possible actions. This means it learns not just how good an action is in a particular state, but also how valuable the state itself is,…
Double Deep Q-Learning
Double Deep Q-Learning is an improvement on the Deep Q-Learning algorithm used in reinforcement learning. It helps computers learn to make better decisions by reducing errors that can happen when estimating future rewards. By using two separate networks to choose and evaluate actions, it avoids overestimating how good certain options are, making learning more stable…
Deep Q-Networks (DQN)
Deep Q-Networks, or DQNs, are a type of artificial intelligence that helps computers learn how to make decisions by using deep learning and reinforcement learning together. DQNs use neural networks to estimate the value of taking certain actions in different situations, which helps the computer figure out what to do next. This method allows machines…
On-Policy Reinforcement Learning
On-policy reinforcement learning is a method where an agent learns to make decisions by following and improving the same policy that it uses to interact with its environment. The agent updates its strategy based on the actions it actually takes, rather than exploring alternative possibilities. This approach helps the agent gradually improve its behaviour through…
Off-Policy Reinforcement Learning
Off-policy reinforcement learning is a method where an agent learns the best way to make decisions by observing actions that may not be the ones it would choose itself. This means the agent can learn from data collected by other agents or from past actions, rather than only from its own current behaviour. This approach…
Conditional Generative Models
Conditional generative models are a type of artificial intelligence that creates new data based on specific input conditions or labels. Instead of generating random outputs, these models use extra information to guide what they produce. This allows for more control over the type of data generated, such as producing images of a certain category or…
Neural Fields
Neural fields are a way to use neural networks to represent and process continuous data, like shapes or scenes, as mathematical functions. Instead of storing every detail as a list of values, neural fields learn to generate the values for any point in space by using a network. This approach can store complex information efficiently…
Geometric Deep Learning
Geometric deep learning is a field of machine learning that focuses on using shapes, graphs, and other complex structures as data instead of just fixed grids like images or text. It allows computers to analyse and learn from data that has relationships or connections, such as social networks, molecules, or 3D shapes. This approach helps…
Equivariant Neural Networks
Equivariant neural networks are a type of artificial neural network designed so that their outputs change predictably when the inputs are transformed. For example, if you rotate or flip an image, the network’s response changes in a consistent way that matches the transformation. This approach helps the network recognise patterns or features regardless of their…