Unlock instant, AI-driven research and patent intelligence for your innovation.

Training and data synthesis and probability inference using nonlinear conditional normalizing flow model

A technology for model data and data synthesis, which is applied in biological neural network models, inference methods, complex mathematical operations, etc., and can solve problems such as inability to accurately model multimodal conditional probability distributions.

Pending Publication Date: 2021-01-19
ROBERT BOSCH GMBH
View PDF0 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Furthermore, while conditional normalization flows exist, e.g. as described in open literature [2], such normalization flows are based on affine (linear) coupling layers which typically do not accurately model complex The multimodal conditional probability distribution of , such as the complex multimodal conditional probability distribution of the example above for pedestrian trajectories given observed pedestrian characteristics

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Training and data synthesis and probability inference using nonlinear conditional normalizing flow model
  • Training and data synthesis and probability inference using nonlinear conditional normalizing flow model
  • Training and data synthesis and probability inference using nonlinear conditional normalizing flow model

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0101] Refer below figure 1 and figure 2 Describes the training of the normalized flow model, then describes the normalized flow model and its training in more detail, and then refers to image 3 describe the comparison of the trained normalized flow model with known conditional affine models, and then refer to Figure 4 and Figure 5 Describe different applications of trained normalized flow models, for example in autonomous vehicles.

[0102] figure 1 A system 100 for training a nonlinear conditional normalized flow model for use in data synthesis or probabilistic inference is shown. System 100 may include an input interface for accessing training data 192 comprising data instances and conditioning data 194 defining conditions on the data instances, and for accessing model data 196 defining normalized flow models, such as described further in this specification. For example, if figure 1 As also illustrated in FIG. 2 , the input interface may consist of a data stor...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The learning of probability distributions of data enables various applications, including but not limited to data synthesis and probability inference. A conditional non-linear normalizing flow model,and a system and method for training said model, are provided. The normalizing flow model may be trained to model unknown and complex conditional probability distributions which are at the heart of many real-life applications. For example, the trained normalizing flow model may be used in (semi)autonomous driving systems to infer what the probability is that a pedestrian is at position x at futuretime t given the pedestrian features c, which may be observed from sensor data, or may be used to synthesize likely pedestrian positions x at future time t given the observed pedestrian features c. This may allow the driving system to determine a route avoiding the pedestrian. Various other applications for the trained normalizing flow model are conceived as well.

Description

technical field [0001] The present invention relates to systems and computer-implemented methods for training standardized flow models for use in data synthesis or probabilistic inference. The present invention further relates to systems and computer-implemented methods for synthesizing data instances using trained normalized flow models, and systems and computer-implemented methods for inferring probabilities of data instances using normalized flow models. The invention further relates to trained normalized flow models. The invention further relates to a computer-readable medium comprising data representing instructions arranged to cause a processor system to perform a computer-implemented method. Background technique [0002] Unknown data probability distributions are at the heart of many real-life problems and can be estimated ("learned") from data using machine learning. Where a probability distribution has been estimated, it is possible to infer probabilities, such as...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): G06N3/08G06N3/04G06N5/04B60W10/18B60W10/20B60W30/09B60W60/00
CPCG06N3/08G06N5/041B60W30/09B60W10/20B60W10/18B60W60/0017B60W2556/35B60W2710/18B60W2710/20G06N3/047G06N3/045G06F18/10G06F17/18G06F18/25
Inventor A·巴塔查里亚C-N·施特雷勒
Owner ROBERT BOSCH GMBH