グラフベースの時間的分類を用いたニューラルネットワークの訓練

Title:

グラフベースの時間的分類を用いたニューラルネットワークの訓練

Document Type and Number:

Japanese Patent JP7466784

Kind Code:

B2

Abstract:

A method for training a neural network with a graph-based temporal classification (GTC) objective function, using a directed graph of nodes connected by edges representing labels and transitions among the labels, is provided. The directed graph specifies one or a combination of non-monotonic alignment between a sequence of labels and a sequence of probability distributions and constraints on the label repetitions. The method comprises executing a neural network to transform a sequence of observations into the sequence of probability distributions, and updating parameters of the neural network based on the GTC objective function configured to maximize a sum of conditional probabilities of all possible sequences of labels that are generated by unfolding the directed graph to the length of the sequence of observations and mapping each unfolded sequence of nodes and edges to a possible sequence of labels.

More Like This:

WO/2020/247200	LIKELIHOOD RATIOS FOR OUT-OF-DISTRIBUTION DETECTION
WO/2023/047117	A COMPUTER-IMPLEMENTED METHOD, DATA PROCESSING APPARATUS, AND COMPUTER PROGRAM FOR ACTIVE LEARNING FOR COMPUTER VISION IN DIGITAL IMAGES
JP6750118	Signal processing method and signal processing device based on spike neural network

Inventors:

Moritz, Nico
Takaaki Hori
Le Lou, Jonathan

Application Number:

JP2023541142A

Publication Date:

April 12, 2024

Filing Date:

July 02, 2021

Export Citation:

Click for automatic bibliography generation Help

Assignee:

Mitsubishi Electric Corporation

International Classes:

G06N3/04; G06N3/047; G10L25/30; G10L25/51

Foreign References:

US7571098

Other References:

GRAVES, Alex et al.,Connectionist temporal classification: labelling unsegmented sequence data with recurrent neural networks,Proceedings of the 23rd International Conference on Machine Learning [online],2006年,pp. 369-376,[検索日2024.02.26], インターネット:
MANGU, Lidia et al.,Finding consensus in speech recognition: word error minimization and other applications of confusion networks,arXiv [online],2000年,pp. 1-35,[検索日2024.02.26], インターネット:
松吉大輝ほか,音響イベント検出におけるBLSTM-CTCを用いた弱ラベル学習の検討,日本音響学会 2018年春季研究発表会講演論文集CD-ROM [CD-ROM],2018年,63-66頁

Attorney, Agent or Firm:

Patent Attorney Fukami Patent Office

Previous Patent: Optical transmitter, optical receiver, optical transmission system, and optical transmission method

Next Patent: Controller with early termination in mixed integer optimal control optimization