W artykule opisano różne techniki stosowane w algorytmach sztucznej inteligencji generatywnej, takie jak modele oparte na rozkładach prawdopodobieństwa, modele wariancyjne oraz modele sekwencyjne. Wyjaśniono podstawy tych technik oraz omówiono ich zastosowania w generowaniu obrazów, muzyki, tekstu czy mowy. Artykuł podkreśla znaczenie algorytmów generative AI jako narzędzi do twórczego generowania treści oraz prezentuje możliwe sposoby zwiększenia efektywności generowania tych treści z wykorzystaniem techniki prompt engineering.
EN
The article describes various techniques used in generative artificial intelligence algorithms, such as models based on probability distributions, variance models and sequential models. The basics of these techniques are explained and their applications in generating images, music, text or speech are discussed. The article emphasizes the importance of generative AI algorithms as tools for creative content generation and presents possible ways to increase the efficiency of generating this content using the prompt engineering technique.
2
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
Artykuł przedstawienia proces budowy modelu generującego wielogłosowe muzyczne sekwencje o określonej emocji. Opisano w nim proces przygotowania bazy przykładów uczących i budowę modelu generatywnego na bazie wariacyjnego autoenkodera. Przedstawiono eksperymenty implementacji warstw konwolucyjnych przeznaczonych do analizy wizualnej reprezentacji przykładów muzycznych. Wygenerowane pliki muzyczne poddano ewaluacji przez użycie metryk i porównanie ze zbiorem treningowym.
EN
This article presents the process of building a system generating polyphonic music content with a specified emotion. The process of preparing a training files and building a generative model based on a variational autoencoder was described. Experiments on the implementation of convolutional layers intended for analysis of the musical examples were presented. The generated examples were evaluated by using metrics and comparing them with the training set.
Material Science is a key factor in the evolution of many industrial sectors. Fields such as the aeronautics, automotive, construction, and biotechnology industries have experienced tremendous development with the introduction of advanced, high-performance materials. Such materials not only provide new functionalities to products, but also significant consequences in terms of economic and environmental sustainability of the products and processes triggered by the more efficient use of energy that they provide. Under this scenario, materials that provide such high performance, such as high entropy alloys (HEAs) or polymer derived ceramics (PDCs), have captured the attention of both industry and researchers in recent years. However, the remarkable number of resources required to develop such materials, from its design phase to its synthesis and characterization, means that the discovery of new high-performance materials is moving at a relatively low pace. This fact places emergent strategies based on artificial intelligence (AI) for the design of materials in a good position to be used to accelerate the whole process, providing an impulse in the initial phases of materials design. The enormous number of combinations of elements and the complexity of synthesizability conditions of HEAs and PDCs respectively, paves the way to the deployment of AI techniques such as Generative Models addressed in this work to create synthetic HEAs and PDCs for highly intensive industrial processes. A specific conditional tabular generative adversarial network (CTGAN) was developed to be used on tabular data to generate novel synthetic compounds for each kind of material. The generated synthetic data was based on the conventional parametric design parameters used for HEAs and PDCs, with specific datasets created for them. The real and generated data are compared, calculation of phase diagrams (CALPHAD) simulations are provided to evaluate the performance of the generated samples and a verification of the novel generated compositions is done in open materials databases available in the literature.
An adaptive and precise peak wavelength detection algorithm for fibre Bragg grating using generative adversarial network is proposed. The algorithm consists of generative model and discriminative model. The generative model generates a synthetic signal and is sampled for training using a deep neural network. The discriminative model predicts the real fibre Bragg grating signal by the calculation of the loss functions. The maxima of loss function of the discriminative signal and the minima of loss function of the generative signal are matched and the desired peak wavelength of fibre Bragg grating is determined. The proposed algorithm is verified theoretically and experimentally for a single fibre Bragg grating peak. The accuracy has been obtained as ±0.2 pm. The proposed algorithm is adaptive in the sense that any random fibre Bragg grating peak can be identified within a short wavelength range.
In this paper we discuss a class of AutoEncoder based generative models based on one dimensional sliced approach. The idea is based on the reduction of the discrimination between samples to one-dimensional case. Our experiments show that methods can be divided into two groups. First consists of methods which are a modification of standard normality tests, while the second is based on classical distances between samples. It turns out that both groups are correct generative models, but the second one gives a slightly faster decrease rate of Fréchet Inception Distance (FID).
6
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
With the increasing technological development in the area of motors, heavy–duty vehicles have been suffering an increase in size and in load capacity. Friction brakes are to decelerate a vehicle by transforming the kinetic energy of the Vehicle to heat, via friction, and dissipating vibration that heat to the surroundings, which produces excessive heat on lining surface. This shows increase in frictional area will definitely reduce the load on brakes by sharing the energy of the vehicle. So the above- factor is taken into account and finally reduction in inertia forces on rotating shaft by providing more frictional area is discussed on this paper. Also give details about applying the frictional force on differential gear shaft. To achieve this inner shoe, which is less than the size of the outer shoe, is provided as per the design of the system developed with the aid of solid works modeling tool. This can be actuated by a specially designed cam, which will actuate both outer shoe and inner shoe respectively. During braking the outer shoe will engage previously to absorb energy in the drum before the inner shoe get actuated. When the cam moves both the shoe remaining energy in both shoes share the vehicles. This entire system is analyzed by using FEA tool ANSYS 14.0 to determine the thermal stress developed in it. These results are compared with the conventional braking system.
W artykule objaśniono pojęcie modelu autogenerującego CAD, jego genezę oraz wynikającą stąd potrzebę budowy tego typu modeli. Krótko omówiono proces tworzenia modelu autogenerującego oraz specyficzne formy zapisu wiedzy stosowane w fazie jego implementacji w różnych systemach CAD. Metodykę budowy modelu autogenerującego przedstawiono na przykładzie zazębienia przekładni ślimakowej, który zrealizowano w oprogramowaniu CATIA. Wskazano źródła i rodzaje wiedzy projektowo-konstrukcyjnej potrzebne do zbudowania ww. modelu oraz język UML, jako metodę formalnego zapisu tej wiedzy. Opisano koncepcję budowy modelu, tj. przyjęte założenia oraz strukturę i logikę jego działania. Przedstawiono również wybrane fragmenty projektu, pokazujące, w jaki sposób model został wykonany.
EN
This article introduces the term of a generative CAD model, its origins and, thus, a need of creating such a type of models. A process of generative model creation as well as specific forms of knowledge recording applied in the implementation phase in various CAD systems are briefly discussed. The example of a worm gear meshing realized by the CATIA software encapsulates the methodology of generative model construction. Sources and types of knowledge for design and construction required for development of the aforementioned model as well as the UML language as a method of formal knowledge recording are presented. The concept of model creation, i.e. assumptions and the structure as well as logic of the model operation are described. Also, the paper addresses selected elements of the project that present the manner in which the model was constructed.
Biologically inspired artificial neural networks have been widely used for machine learning tasks such as object recognition. Deep architectures, such as the Convolutional Neural Network, and the Deep Belief Network have recently been implemented successfully for object recognition tasks. We conduct experiments to test the hypothesis that certain primarily generative models such as the Deep Belief Network should perform better on the occluded object recognition task than purely discriminative models such as Convolutional Neural Networks and Support Vector Machines. When the generative models are run in a partially discriminative manner, the data does not support the hypothesis. It is also found that the implementation of Gaussian visible units in a Deep Belief Network trained on occluded image data allows it to also learn to effectively classify non-occluded images.
9
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
In this paper, we use Sequential Forward Selection to select 8 dimensional frame-level features from the total 69 dimensional features, and we reduce the dimensions of utterance-level eigenvectors from 63 to 12 by fisher discriminant. Then, two kinds of GMM multidimensional likelihoods are proposed for hybrid generative and discriminative models. Experimental results on Berlin emotional speech databases show that the GMM-MAP/SVM series hybrid model is the optimal Hybrid Generative and Discriminative Models, with the recognition rate up to 85.1%.
PL
W artykule zaprezentowano system wykrywania emocji w głosie na podstawie modelu dyskryminacyjnego. Zaprezentowano badania skuteczności system na przykładzie bazy danych Berlin.
The paper present comparison of methods for implementing VBA for creation of KBE application in CATIA system. The usage of VBA is one of the methods of automation of Generative Model creation in CATIA system. Creation of KBE Package forms one of the last phases of the whole process of KBE creation which consists of capturing knowledge necessary for KBE system, formal representation of knowledge, and transferring and implementing knowledge in KBE package. While creating KBE application, Generative Model is being formed and for the purpose of formalization of creation process of that model CATIA Knowledgeware tools are used. In order to have automation of that process it is necessary to ensure programming techniques, including VBA. The suggested 4 methods of implementing VBA for automation of Generative Model creation guarantee greater stability and repeatability of the model.
PL
W artykule przedstawiono porównanie metod zastosowania VBA w końcowym etapie tworzenia aplikacji projektowej opartej na wiedzy (Knowledge Based Engineering - KBE) w systemie CATIA. Zastosowanie VBA jest jedną z metod automatyzacji tworzenia modelu autogenerującego w systemie CATIA, który jest głównym elementem aplikacji KBE. Tworzenie pakietu aplikacji KBE jest jednym z ostatnich etapów całego procesu tworzenia systemu KBE, na który składają się akwizycja wiedzy koniecznej dla systemu KBE, formalna reprezentacja wiedzy i transfer i integracja wiedzy w aplikacji KBE. W fazie tworzenia aplikacji KBE tworzony jest model autogenerujący i dla formalizacji procesu tworzenia tego modelu wykorzystywane są narzędzia Knowledgeware systemu CATIA. Automatyzacja tego procesu wymaga zapewnienia technik programowych w tym zastosowania VBA. Zaproponowane 4 metody implementacji VBA do automatyzacji procesu tworzenia modelu autogenerującego pozwalają na zwiększenie stabilności i powtarzalności tego modelu.
JavaScript jest wyłączony w Twojej przeglądarce internetowej. Włącz go, a następnie odśwież stronę, aby móc w pełni z niej korzystać.