Test Your Knowledge of Artificial Intelligence Concepts

• Intelligence is the ability of a system to act appropriately in an uncertain environment, where appropriate action is that which increases the probability of success, and success is the achievement of behavioral subgoals that support the system’s ultimate goal.
• Symbolic AI is the term for the collection of all methods in artificial intelligence research that are based on high-level human-readable representations of problems, logic and search.
• Reinforcement learning can be defined as a mapping from states to actions, where the agent learns to associate positive and negative reinforcement with specific behaviors.
• The euclidean distance is a measure of similarity between points in euclidean space, and is typically used for comparing vectors.
• Cosine similarity is a measurement that quantifies the similarity between two or more vectors, by measuring the cosine of the angle between them.
• Convolution is the process of applying a filter or a kernel to a signal.
• Batch Normalization is a layer that reduces the variance in the inputs to a layer in a neural network.
• Generalization is the ability of the network to perform on previously unseen data.
• Regularization is a technique that reduces overfitting.
• The key concept in ResNet is the use of skip connections, which allow gradient information to pass through the layers.
• Top 5 Accuracy is preferred metric for ILSVRC, as it allows for a more comprehensive evaluation of the performance of a model.
• LeNet is a deep learning network with 5 layers and 60 million parameters
• LeNet is one of the first successful applications of CNNs
• ReLU’s can die when the input to the ReLU is consistently negative, in which case the output will always be 0 and the ReLU will become inactive
• Keras is an abstraction that sits on top of Tensorflow with a higher layer of abstraction so it can do a lot of the heavy lifting for you
• Regularization helps reduce overfitting, increases generalization
• Value-based approaches vs policy based approaches (a and b mostly)
• Dynamic Programming is a method used to back up plans in case of errors
• Machine Learning is a computer program that learns from experience
• Reinforcement learning is a process of learning what to do in order to maximise a numerical reward signal.
• There are two main types of reinforcement learning: on-policy and off-policy.
• On-policy methods improve the policy that is used to make decisions, whereas off-policy methods improve a policy different from that used to generate the data.
• TD methods are a type of off-policy learning that uses sampling data from real episodes.

In this quiz, you will be asked to identify the main concepts in machine learning. You will also be asked to identify the main types of reinforcement learning. Finally, you will be asked to identify the main benefits of reinforcement learning.