Back to results
US 7,072,875 B2Granted
Information processing apparatus and method, and recording medium
AssigneeSony Corporation
InventorJun Tani
Life of this patent
Estimating the term…
Abstract
An information processing apparatus includes a first recurrent neural network (RNN) for performing processing which corresponds to a time-series and a second RNN for processing another correlated time-series. The difference between a context set output by the first RNN and a context set output by the second RNN is computed by a subtractor, and the obtained difference is used as a prediction error. Backpropagation is performed based on the prediction error, thus determining a coefficient for each neuron of an output layer, an intermediate layer, and an input layer.
Claims1
An information processing apparatus comprising:
- a first recurrent neural network;
- a second recurrent neural network;
- prediction error generating means for generating a prediction error based on a difference between an output of said first recurrent neural network and said second recurrent neural network;
- and
- coefficient determining means for determining a coefficient by performing a backpropagation based on said prediction error in order to minimize said prediction error,
- wherein one of said first recurrent neural network and said second recurrent neural network comprises processing means for performing inverse dynamics processing based on a context output by the other recurrent neural network,
- wherein predetermined speech is input to the second recurrent neural network which performs forward dynamics processing and obtains said context output, and
- said inverse dynamics processing are repeated a predetermined number of times by inputting an output as feedback.
5 more elements
Dependencies are inferred from the claim text.