scholarly journals Commonsense Reasoning to Guide Deep Learning for Scene Understanding (Extended Abstract)

2020 ◽  
Author(s):  
Mohan Sridharan ◽  
Tiago Mota
Keyword(s):  
Domain Knowledge ◽  
Inductive Learning ◽  
Network Models ◽  
Computational Effort ◽  
Partial Occlusion ◽  
Commonsense Reasoning ◽  
Deep Networks ◽  
Training Examples ◽  
The Stability ◽  
Simulated Images

Our architecture uses non-monotonic logical reasoning with incomplete commonsense domain knowledge, and incremental inductive learning, to guide the construction of deep network models from a small number of training examples. Experimental results in the context of a robot reasoning about the partial occlusion of objects and the stability of object configurations in simulated images indicate an improvement in reliability and a reduction in computational effort in comparison with an architecture based just on deep networks.

scholarly journals Integrated Commonsense Reasoning and Deep Learning for Transparent Decision Making in Robotics

2021 ◽  
Vol 2 (4) ◽  
Author(s):  
Tiago Mota ◽  
Mohan Sridharan ◽  
Aleš Leonardis
Keyword(s):  
Decision Making ◽  
Deep Learning ◽  
Domain Knowledge ◽  
Inductive Learning ◽  
Commonsense Reasoning ◽  
Knowledge Based ◽  
New Information ◽  
Robot Systems ◽  
Simulated Images ◽  
Relevant Domain

AbstractA robot’s ability to provide explanatory descriptions of its decisions and beliefs promotes effective collaboration with humans. Providing the desired transparency in decision making is challenging in integrated robot systems that include knowledge-based reasoning methods and data-driven learning methods. As a step towards addressing this challenge, our architecture combines the complementary strengths of non-monotonic logical reasoning with incomplete commonsense domain knowledge, deep learning, and inductive learning. During reasoning and learning, the architecture enables a robot to provide on-demand explanations of its decisions, the evolution of associated beliefs, and the outcomes of hypothetical actions, in the form of relational descriptions of relevant domain objects, attributes, and actions. The architecture’s capabilities are illustrated and evaluated in the context of scene understanding tasks and planning tasks performed using simulated images and images from a physical robot manipulating tabletop objects. Experimental results indicate the ability to reliably acquire and merge new information about the domain in the form of constraints, preconditions, and effects of actions, and to provide accurate explanations in the presence of noisy sensing and actuation.

scholarly journals Templated Text Synthesis for Expert-Guided Multi-Label Extraction from Radiology Reports

2021 ◽  
Vol 3 (2) ◽  
pp. 299-317
Author(s):  
Patrick Schrempf ◽  
Hannah Watson ◽  
Eunsoo Park ◽  
Maciej Pajak ◽  
Hamish MacKinnon ◽  
...  
Keyword(s):  
Domain Knowledge ◽  
Expert Knowledge ◽  
Medical Image Analysis ◽  
Imaging Data ◽  
Data Synthesis ◽  
Domain Specific ◽  
Radiology Reports ◽  
Training Examples ◽  
Analysis Models ◽  
Difficult Cases

Training medical image analysis models traditionally requires large amounts of expertly annotated imaging data which is time-consuming and expensive to obtain. One solution is to automatically extract scan-level labels from radiology reports. Previously, we showed that, by extending BERT with a per-label attention mechanism, we can train a single model to perform automatic extraction of many labels in parallel. However, if we rely on pure data-driven learning, the model sometimes fails to learn critical features or learns the correct answer via simplistic heuristics (e.g., that “likely” indicates positivity), and thus fails to generalise to rarer cases which have not been learned or where the heuristics break down (e.g., “likely represents prominent VR space or lacunar infarct” which indicates uncertainty over two differential diagnoses). In this work, we propose template creation for data synthesis, which enables us to inject expert knowledge about unseen entities from medical ontologies, and to teach the model rules on how to label difficult cases, by producing relevant training examples. Using this technique alongside domain-specific pre-training for our underlying BERT architecture i.e., PubMedBERT, we improve F1 micro from 0.903 to 0.939 and F1 macro from 0.512 to 0.737 on an independent test set for 33 labels in head CT reports for stroke patients. Our methodology offers a practical way to combine domain knowledge with machine learning for text classification tasks.

scholarly journals Mutualistic strategies minimize coextinction in plant–disperser networks

2017 ◽  
Vol 284 (1854) ◽  
pp. 20162302 ◽  
Author(s):  
Evan C. Fricke ◽  
Joshua J. Tewksbury ◽  
Elizabeth M. Wandrag ◽  
Haldre S. Rogers
Keyword(s):  
Network Structure ◽  
Field Experiments ◽  
Empirical Relationship ◽  
Network Models ◽  
Mutualistic Interactions ◽  
Simulated Network ◽  
Direct And Indirect Impacts ◽  
Indirect Impacts ◽  
The Stability ◽  
Seed Dispersal Mutualism

The global decline of mutualists such as pollinators and seed dispersers may cause negative direct and indirect impacts on biodiversity. Mutualistic network models used to understand the stability of mutualistic systems indicate that species with low partner diversity are most vulnerable to coextinction following mutualism disruption. However, existing models have not considered how species vary in their dependence on mutualistic interactions for reproduction or survival, overlooking the potential influence of this variation on species' coextinction vulnerability and on network stability. Using global databases and field experiments focused on the seed dispersal mutualism, we found that plants and animals that depend heavily on mutualistic interactions have higher partner diversity. Under simulated network disruption, this empirical relationship strongly reduced coextinction because the species most likely to lose mutualists depend least on their mutualists. The pattern also reduced the importance of network structure for stability; nested network structure had little effect on coextinction after simulations incorporated the empirically derived relationship between partner diversity and mutualistic dependence. Our results highlight a previously unknown source of stability in mutualistic networks and suggest that differences among species in their mutualistic strategy, rather than network structure, primarily accounts for stability in mutualistic communities.

scholarly journals Information Extraction of Cybersecurity Concepts: An LSTM Approach

2019 ◽  
Vol 9 (19) ◽  
pp. 3945 ◽  
Author(s):  
Houssem Gasmi ◽  
Jannik Laval ◽  
Abdelaziz Bouras
Keyword(s):  
Neural Network ◽  
Domain Knowledge ◽  
Conditional Random Fields ◽  
Short Term Memory ◽  
Network Models ◽  
Relation Extraction ◽  
Entity Recognition ◽  
Feature Engineering ◽  
Neural Network Models ◽  
Feature Based

Extracting cybersecurity entities and the relationships between them from online textual resources such as articles, bulletins, and blogs and converting these resources into more structured and formal representations has important applications in cybersecurity research and is valuable for professional practitioners. Previous works to accomplish this task were mainly based on utilizing feature-based models. Feature-based models are time-consuming and need labor-intensive feature engineering to describe the properties of entities, domain knowledge, entity context, and linguistic characteristics. Therefore, to alleviate the need for feature engineering, we propose the usage of neural network models, specifically the long short-term memory (LSTM) models to accomplish the tasks of Named Entity Recognition (NER) and Relation Extraction (RE). We evaluated the proposed models on two tasks. The first task is performing NER and evaluating the results against the state-of-the-art Conditional Random Fields (CRFs) method. The second task is performing RE using three LSTM models and comparing their results to assess which model is more suitable for the domain of cybersecurity. The proposed models achieved competitive performance with less feature-engineering work. We demonstrate that exploiting neural network models in cybersecurity text mining is effective and practical.

Quaternionic Neural Networks

2009 ◽  
pp. 411-439 ◽  
Author(s):  
Teijiro Isokawa ◽  
Nobuyuki Matsui ◽  
Haruhiko Nishimura
Keyword(s):  
Neural Networks ◽  
Dimensional Space ◽  
Three Dimensional ◽  
Network Models ◽  
Recurrent Network ◽  
Affine Transformations ◽  
Neural Network Models ◽  
Fundamental Properties ◽  
Modern Physics ◽  
The Stability

Quaternions are a class of hypercomplex number systems, a four-dimensional extension of imaginary numbers, which are extensively used in various fields such as modern physics and computer graphics. Although the number of applications of neural networks employing quaternions is comparatively less than that of complex-valued neural networks, it has been increasing recently. In this chapter, the authors describe two types of quaternionic neural network models. One type is a multilayer perceptron based on 3D geometrical affine transformations by quaternions. The operations that can be performed in this network are translation, dilatation, and spatial rotation in three-dimensional space. Several examples are provided in order to demonstrate the utility of this network. The other type is a Hopfield-type recurrent network whose parameters are directly encoded into quaternions. The stability of this network is demonstrated by proving that the energy decreases monotonically with respect to the change in neuron states. The fundamental properties of this network are presented through the network with three neurons.

Complex Network Model and its Application

2013 ◽  
Vol 791-793 ◽  
pp. 1589-1592
Author(s):  
Shuai Xu ◽  
Bai Da Zhang
Keyword(s):  
Complex Network ◽  
Road Network ◽  
Human Life ◽  
Network Models ◽  
Physical Object ◽  
Statistical Characteristics ◽  
Power Network ◽  
The Stability ◽  
Complex Network Models ◽  
Important Significance

Human life is in a complex network world. In everyday life, the network can be a physical object such as the Internet, power network, road network and neural network; can also abstract not touch, such as interpersonal networks, networks of co-operation in scientific research, product supply chain network, biological populations, networks, etc.. The topology of these networks, the statistical characteristics and the formation mechanism, and so on, has a very important significance for the efficient allocation of resources, provides various functions, as well as the stability of the network, however, due to the complexity of these networks, conventional simplified model and cannot be good solution to the above problems. The complex network and network complexity has become a hot issue in the scientific and engineering concern. This article describes a few common complex network models and its application brief.

Multifunctional Electrical Prosthetic Hand -Development of Tendon-driven Mechanism and Controller

2002 ◽  
Vol 14 (6) ◽  
pp. 557-564 ◽  
Author(s):  
Wenwei Yu ◽  
◽  
Daisuke Nishikawa ◽  
Yasuhiro Ishikawa ◽  
Hiroshi Yokoi ◽  
...  
Keyword(s):  
Neural Network ◽  
Time Delay ◽  
Recurrent Neural Network ◽  
Network Models ◽  
Joint Torque ◽  
Prosthetic Hand ◽  
Neural Network Models ◽  
Nonlinear Properties ◽  
The Stability

The purpose of this research was to develop a tendondriven electrical prosthetic hand, which is characterized by its mechanical torque-velocity converter and a mechanism that can assist proximal joint torque by distal actuators. To cope with time-delay and nonlinear properties of the prosthetic hand, a controller based on a Jordan network, recurrent neural network models, is proposed. The results of experiments on the stability of the controller are confirmed when tracking static wire tensions. Finally, the next prototype of prosthetic hand based on these methods is introduced.

Speed Regulation of a Permanent Magnet Synchronous Motor via Model Reference Adaptive Control

2011 ◽  
Vol 268-270 ◽  
pp. 513-516
Author(s):  
Zhi Yong Qu ◽  
Zheng Mao Ye
Keyword(s):  
Permanent Magnet ◽  
Permanent Magnet Synchronous Motor ◽  
Estimation Algorithm ◽  
Synchronous Motor ◽  
Computational Effort ◽  
Model Reference ◽  
Speed Regulation ◽  
The Stability ◽  
Stator Resistance ◽  
Model Reference Adaptive

A speed estimation technique for the permanent magnet synchronous motor drive is presented in this paper A Model Reference Adaptive System (MRAS) has been formed using the voltage and current to estimate the speed. It has been shown that such unique MRAS offers several desirable features. The proposed technique is completely independent of stator resistance and is less parameter sensitive, as the estimation-algorithm is only dependent on q-axis stator inductance. Also, the method requires less computational effort as the simplified expressions are used in the MRAS. The stability of the proposed system is achieved through Popov’s Hyperstability criteria. Matlab simulation results are presented to validate the proposed technique.

scholarly journals PyInteraph2 and PyInKnife2 to analyze networks in protein structural ensembles

2020 ◽  
Author(s):  
Valentina Sora ◽  
Matteo Tiberti ◽  
Shahriyar Mahdi Robbani ◽  
Joshua Rubin ◽  
Elena Papaleo
Keyword(s):  
Network Models ◽  
Version Control ◽  
Complex Nature ◽  
Cellular Functions ◽  
Diverse Range ◽  
Link Type ◽  
Network Properties ◽  
Protein Ensembles ◽  
The Stability ◽  
Definition Of

AbstractMotivationProtein dynamic is essential for cellular functions. Due to the complex nature of non-covalent interactions and their long-range effects, the analysis of protein conformations using network theory can be enlightening. Protein Structure Networks (PSNs) rely on different philosophies, and the currently available tools suffer from limitations in terms of input formats, supported network models, and version control. Another issue is the precise definition of cutoffs for the network calculations and the assessment of the stability of the parameters, which ultimately affect the outcome of the analyses.ResultsWe provide two open-source software packages, i.e., PyInteraph2 and PyInKnife2, to implement and analyze PSNs in a harmonized, reproducible, and documented manner. PyInteraph2 interfaces with multiple formats for protein ensembles and calculates a diverse range of network models with the possibility to integrate them into a macro-network and perform further downstream graph analyses. PyInKnife2 is a standalone package that supports the network models implemented in PyInteraph2. It employs a jackknife resampling approach to estimate the convergence of network properties and streamline the selection of distance cutoffs. Several functionalities are based on MDAnalysis and NetworkX, including parallelization, and are available for Python 3.7. PyInteraph2 underwent a massive restructuring in terms of setup, installation, and test support compared to the original PyInteraph software.ConclusionsWe foresee that the modular structure of the code and the version control system of GitHub will promote the transition to a community-driven effort, boost reproducibility, and establish harmonized protocols in the PSN field. As developers, we will guarantee the introduction of new functionalities, assistance, training of new contributors, and maintenance of the package.AvailabilityThe packages are available at https://github.com/ELELAB/pyinteraph2 and https://github.com/ELELAB/PyInKnife2 with guides provided within the packages.

EXPLORING THE VULNERABILITY OF FRACTAL COMPLEX NETWORKS THROUGH CONNECTION PATTERN AND FRACTAL DIMENSION

Fractals ◽  
2019 ◽  
Vol 27 (06) ◽  
pp. 1950102
Author(s):  
DONG-YAN LI ◽  
XING-YUAN WANG ◽  
PENG-HE HUANG
Keyword(s):  
Fractal Dimension ◽  
Quantitative Research ◽  
Fractal Dimensions ◽  
Network Models ◽  
Small World ◽  
Brain Network ◽  
Complex Model ◽  
Scale Free ◽  
Connection Pattern ◽  
The Stability

The structure of network has a significant impact on the stability of the network. It is useful to reveal the effect of fractal structure on the vulnerability of complex network since it is a ubiquitous feature in many real-world networks. There have been many studies on the stability of the small world and scale-free models, but little has been down on the quantitative research on fractal models. In this paper, the vulnerability was studied from two perspectives: the connection pattern between hubs and the fractal dimensions of the networks. First, statistics expression of inter-connections between any two hubs was defined and used to represent the connection pattern of the whole network. Our experimental results show that statistic values of inter-connections were obvious differences for each kind of complex model, and the more inter-connections, the more stable the network was. Secondly, the fractal dimension was considered to be a key factor related to vulnerability. Here we found the quantitative power function relationship between vulnerability and fractal dimension and gave the explicit mathematical formula. The results are helpful to build stable artificial network models through the analysis and comparison of the real brain network.

Sign in / Sign up

Export Citation Format

Share Document