Dave's PCF WIP: Paragraphs

Paragraph #15,474

Edit Paragraph

View List Paragraphs Delete

Application

Matter Number

Paragraph Number

Application Number

Reference Case 1

Reference Case 2

Content

Artificial Intelligence can utilize, for example, neural networks. Artificial neural networks (ANNs) can learn tasks based on examples, e.g. without task specific programming. ANNs can be based on a group of connected units or nodes, i.e. artificial neurons. Each connection between artificial neurons can generate a signal to be transmitted to another artificial neuron. One or more artificial neurons can receive the signal and process it and, for example, use it to initiate a task. The signal at a connection between artificial neurons can be a number, and the outputs can be calculated by various functions or algorithms, e.g. non-linear functions. Artificial neurons can have a weight assigned to them, which can amplify or de-emphasize their signal. Artificial neurons can be organized in layers, with different layers performing different kinds of transformations. Artificial neural networks can utilize various techniques, processes and/or algorithms, e.g. backpropagation, parallel distributed processing, max-pooling, Hebbian learning, long term potentiation, support vector machines, and linear classifiers. ANNs can include recurrent neural networks and deep feedforward neural networks. ANNs can perform functions such as, for example, pattern recognition and machine learning. Components of ANNs can include neurons, connections and weights, propagation functions, and learning rules. Neurons can include an activation component, a threshold component, an activation function, and an output function. ANNs can define mathematical and other functions. ANNs can use predefined functions, e.g. hyperbolic tangent function, sigmoid function, softmax function or rectifier function.

Notes

Raw Data