Greece

Common types of artificial neural networks have been well known to suffer from the presence of outlying measurements (outliers) in the data. However, there are only a few available robust alternatives for training common form of neural networks. In this work, we investigate robust fitting of multilayer perceptrons, i.e. alternative approaches to the most common type of feedforward neural networks. Particularly, we consider robust neural networks based on the robust loss function of the least trimmed squares, for which we express formulas for 
derivatives of the loss functions. Some formulas, which are however incorrect, have been already available. Further, we consider a very recently proposed multilayer perceptron based on the loss function of the least weighted squares, which appears a promising highly robust approach. We also derive the derivatives of the loss functions, which are to the best of our knowledge a novel contribution of this paper. The derivatives may find applications in implementations of the robust neural networks, if a (gradient-based) backpropagation algorithm is used.

EANN 2020

Robust multilayer perceptrons: Robust loss functions and their derivatives

robust training

outliers

neural networks

The purpose of the AAAI conference series is to promote research in Artificial Intelligence (AI) and foster scientific exchange between researchers, practitioners, scientists, students, and engineers across the entirety of AI and its affiliated disciplines. AAAI-23 is the Thirty-Seventh AAAI Conference on Artificial Intelligence. The theme of this conference is to create collaborative bridges within and beyond AI. Like previous AAAI conferences, AAAI-23 will feature technical paper presentations, special tracks, invited speakers, workshops, tutorials, poster sessions, senior member presentations, competitions, and exhibit programs, and two new activities: a Bridge Program and a Lab Program. Many of these activities are tailored to the theme of bridges and all are selected according to the highest standards, with additional programs for students and young researchers. 
AAAI is providing you with a conference planner, which you can use to help organize your itinerary of activities. This includes talks to attend in person, talks to attend remotely, breaks with colleagues and your site seeing activities. To access this conference planner, please go to [https://aaai-2023.takemobi.io](https://aaai-2023.takemobi.io).

In order to access this site, you need to register. If you haven't already, please register [here](https://aaai.org/Conferences/AAAI-23/registration/).


AAAI 2023

The purpose of the AAAI conference series is to promote research in Artificial Intelligence (AI) and foster scientific exchange between researchers, practitioners, scientists, students, and engineers across the entirety of AI and its affiliated disciplines.

technical paper

21st edition of EANN (International Conference on Engineering Applications of Neural Networks) should be held in Halkidiki, Greece. Due to current Covid-19 situation the conference will be held virtually and Underline will host a repository of video talks and preserve them for the future. Videos will be released on Underline one day after the conference. 

A fully-connected Deep Neural Network (DNN) architecture, to be referred to as λ-DNN, used to predict 2D/3D scalar fields is presented. In aerodynamics, the λ-DNN is firstly trained on fields computed using a Computational Fluid Dynamics (CFD) software. Then, it can be incorporated into engineering processes in various ways. One possibility is to use them in optimization problems solved by stochastic population-based methods, in which the λ-DNN may act as the surrogate evaluation model, replacing calls to the CFD tool. Another possibility is in multi-disciplinary problems, to replicate the numerical solver for any of the disciplines. This small list of possible usages is not exhaustive and, of course, different usages can be combined. The input to each DNN contains information to identify the geometrical shape and case-related data, nodal coordinates and, in multi-disciplinary problems, interfacing data connecting solvers for different disciplines on adjacent domains. In this presentation, the λ-DNN is firstly used in the aerodynamic shape optimization of a wing using evolutionary algorithms, in which it replicates the CFD solver. Then, it is used in a conjugate heat transfer problem dealing with a solid domain in contact with a flow within a duct. In this problem, the λ-DNN acts as a surrogate to the solver of the heat conduction equation on the solid domain, by interfacing with a CFD solver of the fluid domain.

λ-DNNs and their Implementation in Aerodynamic and Conjugate Heat Transfer Optimization

The combinatorial pharmacological effects of drugs are influenced by the complex interactions among them. When the patient is allergic to some drugs, the combination of drugs has to be changed. Based on that, we put forward a linear-algebra-equation query task. In this paper, we propose a model called Drug2vec that approximates the relationship among drugs and can solve the linear-algebra-equation query task. For example, we can find drugs with the following relationship: drug A + drug B = drug C. Drug2vec applies a three-layer neural network, which firstly projects a drug into an embedded space and then retrieves another drug that interacts with it. Experimental results show that Drug2vec can approximate the relationship among drugs to linear equations, and the drugs that fit a linear equation have connections with respect to their structures. We also propose a metric called AUE (area under the enrichment curve) to evaluate the performance of our model. Drug2vec can predict drug-drug interactions with high accuracy, and the AUE can be 0.96 in the normal test.

Drug2vec: A Drug Embedding Method with Drug-Drug Interaction as the Context

We present a project to design of the simplified model of thinking located in a little-explored field between neurobiology and psychology. While neurobiology in the incredibly complex microscopic system is dedicated to the structures and signals at the molecular level and the psychology is committed at the opposite extreme with highly sophisticated, they are very abstract and, therefore, difficult to grasp these wholes. There is a vast space between these two extremes. From the perspective of the inherent observers, it is investigable without overcoming the complexity of both points of interest. The primary goal of this research is to construct a multi-layer analog configuration space, the Electronic Equivalent of Consciousness (ECC), wherein the signals have the same properties (bioelectrical) as in the human brain.

https://screencast-o-matic.com/watch/cYhTcHpfFw"

Electronic equivalent of consciousness with the elementary mental process model

In this presentation, a method is proposed in order to implement a virtual sensor which is able to estimate the evaporated water in cooling towers, considering the variations of environmental and operating conditions. A generative adversarial network (GAN) is used for training a deep learning-based generative model which is able to generalize the estimation of the evaporated water, even in situations not included in the training data. The proposed method is tested using real data from a cooling tower located at the Hospital of León.

Probabilistic estimation of evaporated water in cooling towers using a generative adversarial network

Reduction of data variables is an important issue and it is needed for the processing of higher dimensional data in the application domains and AI, in which threshold neural networks are extensively used. We develop a reduction of data variables and classification method based on the nearest neighbor relations for threshold networks. First, the nearest neighbor relations are shown to be useful for the generation of threshold functions and Chow parameters. Second, the extended application of the nearest neighbor relations is developed for the reduction of variables based on convex cones. The edges of convex cones are compared for the reduction of variables. Further, hyperplanes with reduced variables are obtained on the convex cones for data classification.

Reduction of Variables and Generation of Functions through Nearest Neighbor Relations in Threshold Networks

In the future power system called Smart Grid, power generators using renewable energies will be widely introduced to the power grid and required to balance supply and demand on the grid quickly. Therefore, fast automated demand response (FastADR) that contributes to balance the power grid from demand side through electrical facilities like building air conditioner are focused recently. When electric grid operator will pay incentive to aggregators of the demand side, it is important to estimate accurate baseline load. However, the FastADR must returns quick response by unit of seconds or minute (fine-granularity), therefore it is difficult to estimate baseline load accurately using conventional method. In this research, the baseline load estimation model for air-con time-series data is constructed using long short-term memory (LSTM) neural network, and compared with multilayer perceptron (MLP) neural network model for baseline load estimation. The training and evaluating time-series data is generated by air-con simulator (AE) carried out on the virtual building. In the estimation results using data that were simulated for a month, the average estimation error of the LSTM model is 2.7% and of the MLP model is 5.3%. Therefore, the LSTM model is more effective for baseline estimation than the MLP model. However, data in various situations are required.

LSTM Neural Network for Fine-Granularity Estimation on Baseline Load of Fast Demand Response

Engineering drawings are common across different domains such as Oil \& Gas, construction, mechanical and other domains. Automatic processing and analysis of these drawings is a challenging task. This is partly due to the complexity of these documents and also due to the lack of dataset availability in the public domain that can help push the research in this area. In this paper, we present a multiclass imbalanced dataset for the research community made of 2432 instances of engineering symbols. These symbols were extracted from a collection of complex engineering drawings known as Piping and Instrumentation Diagram (P&ID). By providing such dataset to the research community, we anticipate that this will help attract more attention to an important, yet overlooked industrial problem, and will also advance the research in such important and timely topics. We discuss the datasets characteristics in details, and we also show how Convolutional Neural Networks (CNNs) perform on such extremely imbalanced datasets. Finally, conclusions and future directions are discussed.

Symbols in Engineering Drawings (SiED): An Imbalanced Dataset Benchmarked by Convolutional Neural Networks

Face Recognition (FR) is an important area in computer vision with many applications such as security and automated border controls. The recent advancements in this domain have pushed the per- formance of models to human-level accuracy. However, the varying condi- tions in the real-world expose more challenges for their adoption. In this paper, we investigate the performance of these models. We analyze the performance of a cross-section of face detection and recognition models. Experiments were carried out without any preprocessing on three state- of-the-art face detection methods namely HOG, YOLO and MTCNN, and three recognition models namely, VGGface2, FaceNet and Arcface. Our results indicated that there is a significant reliance by these methods on preprocessing for optimum performance.

Towards a Reliable Face Recognition System

In this paper, we introduce our efficient simple method which can locate all occurrences of pattern P of k subpatterns with “don’t cares” of length m in text S of length n. Our algorithm employs advanced data structure and the Kangaroo method, which can be applied to selected suffixes of the suffix tree of S to answer subsequent queries in O(k) time using a predefined computational method to find all occurrences of pattern P with “don’t cares” in text S in a fast and effective manner.

Detecting Pattern Efficiently with Don’t Cares

Knowledge of permeability, a measure of the ability of rocks to allow fluids to 
flow through them, is essential for building accurate models of oil and gas reservoirs. Permeability is best measured in the laboratory using special core analysis (SCAL), but this is expensive and time-consuming. This is the first major work on predicting permeability in the in the UK Continental Shelf (UKCS) based only on routine
core analysis (RCA) data and a machine-learning approach. We present a
comparative analysis of the various machine learning algorithms and validate
the results, using permeability measured on 273 core samples from
104 wells. Results suggest that machine learning can predict permeability with relatively high accuracy. This opens new research directions in
particular in the oil and gas sector.

Robust multilayer perceptrons: Robust loss functions and their derivatives

Downloads

Next from EANN 2020

λ-DNNs and their Implementation in Aerodynamic and Conjugate Heat Transfer Optimization

Similar lecture

Robust Training of Neural Networks against Bias Field Perturbations

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES

Robust multilayer perceptrons: Robust loss functions and their derivatives

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from EANN 2020

λ-DNNs and their Implementation in Aerodynamic and Conjugate Heat Transfer Optimization

Similar lecture

Robust Training of Neural Networks against Bias Field Perturbations

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES

Downloads