Search 250 million research papers

Read the abstract, see who wrote it and who cites it, and get the citation in APA, MLA, Chicago or Harvard.

About 10,000 papers for “Generative model”

Generative Model

Peng Long, Xiaozhou Guo · 2026

Cited by 0

Generative model

John Darby, Baihua Li, Nicholas Costen · 2010

This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and education use, including for instruction at the authors institution and sharing with colleagues. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier’s archiving and manuscript policies are encouraged to visit:

Cited by 0

Generative model.

Rosalyn J. Moran (209988), Nicolas Mallet (209990), Vladimir Litvak (42646) et al. · Figshare · 2015

This figure summarises the generative model as a Bayesian Network (upper left insert), which has been unpacked to show the form of the conditional dependencies in terms of the equations of the generative model (lower right). This model provides a probabilistic description of data in terms of the parameters that cause them. These include parameters controlling the spectral density of neuronal and channel noise and parameters controlling the variance of observation error on the cross-spectra predicted. The parameters of interest pertain to a neuronal model that is cast as a continuous-time state-space model. It is this neuronal model that constrains the mapping from neuronal noise or innovations to observed cross-spectra. The details of the neuronal (mass) model are provided in the next figure. Please see the main text for a detailed explanation of the equations that define the conditional dependencies.

Cited by 0Open access

Generative Model

Biswadip Dey, P. S. Krishnaprasad, Natural Frenet Frame · 2013

To explore underlying strategies and motion (pursuit, collective motion etc.) governing control laws, by extracting parameters of motion (namely curvature, speed, lateral acceleration etc.) from sampled observations of trajectories. To extract control inputs from sampled data. Trajectory Smoothing as a Linear Optimal Control Problem Problem

Cited by 0

Generative Model Evaluation

George Stein · TIB Data Manager · 2024

Cited by 0Open access

The generative model.

David Benrimoh (6919352), Thomas Parr (5956196), Rick A. Adams (7289171) et al. · Figshare · 2019

This schematic illustrates the specification of the (partially observed Markov decision process) generative model used in the simulations. There are three sorts of hidden state: those that determine the content of speech (i.e. the words in the song), whether the agent is currently listening to anything, and whether the agent is speaking. The first deterministically transitions from each word to the next in the song, and repeats this sequence indefinitely as shown in the B-matrix for this content state Bc. The latter two hidden state factors depend upon the agent’s choice; with actions given in small numbered circles. The policies are defined such that, when action 1 (Speaking) is taken for the listening state (leading to ‘I am not listening’), action 1 is also taken for the speaking state (leading to ‘I am speaking’), and similarly for action 2 (Listening). There is no such constraint in the generative process, so this belief can be violated. When listening, the (lexical) content states map to auditory outcomes representing the same words. This is shown in the first of the auditory A-matrices Aa for which columns represent content states, and rows content outcomes. The second (lower) version of this matrix is employed when not listening to anything and gives rise to a ‘silent’ auditory outcome (7th row). There is additionally a proprioceptive outcome modality op that depends only upon whether the agent is speaking. The equation at the bottom of this Figure shows how we equip the auditory likelihood with a precision (inverse temperature) parameter ζ. The bar above the A indicates that the matrix has been normalised, following precision modulation.

Cited by 0Open access

Simulation generative model.

Jean-Baptiste Pingault (148638), Frühling Rijsdijk (5196188), Tabea Schoeler (7808963) et al. · Figshare · 2021

Cited by 0Open access

Generative model transformer

Jorn Bettin, Ghica van Emde Boas · 2003

The Generative Model Transformer (GMT) project is an Open Source initiative to build a Model Driven Architecure™ tool that allows fully customisable Platform Independent Models, Platform Description Models, Texture Mappings, and Refinement Transformations. The project should result in (a) a tool that fulfils the MDA promise for faster/more accurate/better maintainable application development, (b) a tool for industrial use, and (c) MDA related research-which is encouraged and needed. A distinctive feature of GMT is the emphasis of model transformations as "first-class model citizens". The implementation of model transformations is envisaged to be in conformance with the future OMG modeling standard for Queries, Views, and Transformations (QVT).

Cited by 1

Controlling Poisson Flow Generative Model: Implementing a class conditional generative model

GEORGIADES, IOANNIS (author) · Research Repository (Delft University of Technology) · 2023

With the following paper we are planning to present and explore the possibilities of the the newly introduced Poisson Flow Generative Model (PFGM). More specifically, this work aims to introduce the Conditional Poisson Flow Generative Model (CoPFGM), which by extending the existing repository of the PFGM, it will be able to be trained in a way that allows for conditional image sampling. The work aims to provide a more modular solution that can be easily adjusted for multiple data sets, including custom, as well as datasets taken directly from large Python libraries such as PyTorch and TensorFlow. Our proposed CoPFGM consists of two steps: (i) modifying the input of underlying UNet and (ii) modifying the loss function. More specifically, for (i) we have augmented the input channels of every image with one-hot-like class conditional images, and about (ii) we are introducing an updated loss function which incorporates the Cross-Entropy Loss of the generated images during training. The proposed model is tested against 2 datasets, the MNIST, and the Dilbert dataset, the latter, consists of 1100 custom images of the faces of 6 characters taken from the Dilbert Comic-Strip. The proposed model will be tested and presented in the form of an Ablation Study, with which, we show the conditional behavior of the channel augmentation, and the image improvement in terms of class representation with the Cross-Entropy loss.

Cited by 0Open access

A generative mobility model

Kreuger, Per · KTH Publication Database DiVA (KTH Royal Institute of Technology) · 2021

Many applications in the mobile radio network domain employ simulations to explore e.g. parameter configurations, robustness of protocols and buffer allocation algorithms. User mobility is (together with traffic and radio propagation models), one of the main components of such simulations, and has large impact on e.g. load distribution, cell handover frequency, signal fading and interference. In many simulations, detailed user mobility is as crucial as the physical infrastructure, where the exact position affect fading and reflections, but in others, e.g. load balancing, handover management and radio resource scheduling, coarser models are often sufficient. But even in these cases, properties of the trajectories of individual users will affect the results of the simulation, e.g. where distribution of positions, rest times, and displacement magnitudes and velocities, need to be considered.

Cited by 0Open access

Semi-Implicit Generative Model

Mingzhang Yin, Mingyuan Zhou · arXiv (Cornell University) · 2019

To combine explicit and implicit generative models, we introduce semi-implicit generator (SIG) as a flexible hierarchical model that can be trained in the maximum likelihood framework. Both theoretically and experimentally, we demonstrate that SIG can generate high quality samples especially when dealing with multi-modality. By introducing SIG as an unbiased regularizer to the generative adversarial network (GAN), we show the interplay between maximum likelihood and adversarial learning can stabilize the adversarial training, resist the notorious mode collapsing problem of GANs, and improve the diversity of generated random samples.

Cited by 2Open access

Generative Model for Antibodies

Pereira, Paul · Zenodo (CERN European Organization for Nuclear Research) · 2025

Cited by 0Open access

ŠAUMJAN'S APPLICATIONAL GENERATIVE MODEL

Eugene Verstraelen · Linguistics · 1972

Cited by 0

Generative model based vision

Arthur E.C. Pece, Rasmus Larsen · Computer Vision and Image Understanding · 2007

Cited by 3Open access

Generative network model structure.

Zhaodan Gu (17008494), Zhilei Wu (9210793), Ning Dai (403092) · Figshare · 2023

The encoder and decoder have a stride of 2, and the convolutional kernel size is 4×4.

Cited by 0Open access

The Integrative Generative Model

Abdullah, Basil Yahya · Zenodo (CERN European Organization for Nuclear Research) · 2025

Title: The Integrative Generative Model: A Unified Mathematical Framework for Function Approximation, Pattern Generation, and Autonomous Scientific DiscoveryAuthor: Basil Yahya AbdullahDescription:This work introduces the Integrative Generative Model (IGM), a novel mathematical framework designed to bridge the gap between the expressive power of modern Artificial Intelligence and the transparency of classical scientific methods. The model addresses longstanding challenges in function approximation, particularly for non-smooth and discontinuous phenomena, and provides a new pathway for solving complex inverse problems.The model is founded on the first-principles concept that any continuous phenomenon f(x) can be generated by the convolution of a primary stimulation signal s(x) (representing discrete or continuous causes) and a system-specific response kernel K_n(x) (representing the governing physical law).The core innovation lies in the definition of this kernel, K_n(x), as the derivative of a Generalized Sigmoid Function (GSF). Its behavior is governed by a complex-valued exponent, n = a + bi, which acts as a master control parameter:The real part, a, controls the sharpness of the response, allowing it to model phenomena ranging from smooth (logistic) to impulsive (delta-like).The imaginary part, b, introduces a rotational phase, enabling the generation of complex vortical and helical patterns crucial for modeling phenomena in fluid dynamics, optics, and quantum mechanics.This framework elegantly demonstrates that discrete summation and continuous integration are merely two special cases of this single generative process, thereby unifying the quantum (discrete) and classical (continuous) perspectives.Furthermore, the IGM provides a robust basis for solving ill-posed inverse problems. We detail a regularized deconvolution method for stably inferring the latent stimulation s(x) when the system kernel K_n(x) is known. The paper culminates in tackling the blind deconvolution problem, where both the cause s(x) and the system law K_n(x) are unknown. To solve this, we propose a "Generative Explorer," an AI agent based on a physics-informed Wasserstein GAN with Gradient Penalty (WGAN-GP), capable of discovering both the latent causes and the governing physical laws directly from observed data.The manuscript provides a full mathematical derivation of the model, analysis of its properties, and computational experiments validating its capabilities in three distinct scientific domains: (1) simulating non-linear optical phenomena like self-focusing, (2) decomposing fluid turbulence into its fundamental modes, and (3) recovering latent neural signals from EEG data.Ultimately, the IGM is presented not just as a tool, but as a blueprint for a new class of AI systems capable of performing autonomous scientific discovery.

Cited by 0Open access

ABCDE of generative model.

M. Berk Mirza (4739337), Rick A. Adams (124147), Christoph Mathys (625718) et al. · Figshare · 2018

The upper panel shows the likelihood matrices for the top left quadrant (location 2), under the vertical (but not horizontal) spatial transformation. The likelihood matrices encoded the probability of outcomes given (the four dimensions of) hidden states. The middle panel shows the action-dependent state transition matrices. In this scene construction task, only the sampling location (where) hidden state is action-dependent. The other transition matrices associated with the context and spatial transformations are identity matrices. The action-dependent transition matrix maps to the sampling locations associated with each action (for the first eight policies). However, this mapping changes in the case of a fixed-form policy, which corresponds to a repeating the ninth action in the MDP model. The mapping between sampling locations for the reading-like policy is shown on the rightmost side of the middle panel. The lower panel shows two priors. Firstly, the prior preferences over the what and where outcome modalities in the first six saccades are shown. Here, the columns of the matrices show the preferences (or utilities) over successive time steps; whereas the rows designate the outcomes (six cues under the what modality and eight locations under the where modality). The utilities in the first time step are zero (shown with the white colour), under both modalities; since the sampled location in the first time step is the central fixation. Different shades of grey indicate the absolute value (intensity) of the utilities, where the darker shades are associated with higher utilities. The prior preference matrix under the what modality equips the agent with beliefs that it expects to categorize correctly. The increasing utility over the columns in the prior preference for the where modality means that the tardy sampling (i.e., being undecided) becomes costly. Plus and minus signs indicate the valence of the utilities. These utilities specify the imperatives or instructional set of the empirical game (described below); in terms of incentives or prior preferences. For example, considering the prior preference matrix under the what modality, a correct categorisation is implicitly rewarded with two points and an incorrect categorisation would be penalised with four points. The prior preferences over policies are shown at the bottom. One can define a propensity for a policy in the vector E, which encodes the prior preferences over the policies. An example E is shown at the bottom. The utility of the ninth policy (heuristic strategy, Eh) is defined as log(2), relative to a value of zero for the remaining (eight) policies. The first eight policies correspond to the policies that take the agent to the locations associated with the central fixation, four quadrants and three choice locations. This renders the ninth policy ≈ 3 times more likely.

Cited by 0Open access

Dynamic network generative model

Chayant Tantipathananan, Tanya Berger-wolf · 2013

In this work we present a statistical model for generating realistic dynamic networks over time. Modeling such networks is necessary both to better understand the underlying dynamics of the population, as well as, for generating synthetic networks that emulate the real–world properties, for validating analysis. Classically there are two main schools of network modeling. The approach primarily used in social sciences is to treat the entire network as a maximum likelihood object from a statistical distribution where some of the parameters, such as the number of dyads or triads, are fixed [4]. A radically different approach originates from the random graph community, where generative models are designed to emulate large scale global statistical behaviors in networks, such as the degree distribution [1, 2] and average distance [5], among others. A significant shortcoming of these classes of models is that they are fundamentally evolutionary but not truly dynamic. That is, once a connection is established it cannot be removed. This intuitively contradicts how social links are formed, reinforced, and lost in real world. These models essentially give a “static” representation of the dynamics of interactions aggregated up to a certain point in time. Here is a simple example of what the above mentioned models fail to capture. At given point in time, society exists as a collection of loosely formed communities [3, 7]. As an individual joins a community, she/he forms relationships with its members. Overtime the individual updates those relationships by adding more links to already existing members or new–comers and by removing other links. Moreover, some individuals leave the communities all–together. Most of the models in the aforementioned two classes capture one or the other network properties but fail to incorporate many others. Especially, the existing models do not capture dynamic of forming and breaking relationships in a fluid community membership. 1

Cited by 0

Generative model for HTM.

Dileep George (258470), Jeff Hawkins (258476) · Figshare · 2015

Hierarchical Temporal Memory (HTM) is a model of neocortical function. HTMs can be specified using a generative model. Shown is a simple two-level three-node HTM-type generative model. Each node in the hierarchy contains a set of coincidence patterns (labeled with ) and a set of Markov chains (labeled with ) defined over the set of coincidence patterns. A coincidence pattern in a node represents a co-activation of particular Markov chains of its child nodes. HTM generative model is a spatio-temporal hierarchy in which higher levels remain stable for longer durations of time and can generate faster changing activations in lower levels.

Cited by 0Open access

Tasks and generative model.

Yanli Zhou (725617), Luigi Acerbi (118009), Wei Ji Ma (9707948) · Figshare · 2020

A: Collinearity judgment task. After stimulus offset, participants reported if the line segments belonged to the same line or different lines. B: Height judgment task. Participants reported whether the left line segment was higher or the right line segment was higher. C: Generative model of the collinearity judgment task. Trial type C = 1 when the two lines segments are collinear, and C = 0 when line segments are non-collinear. On a given trial, the stimulus pair yL, yR randomly appeared around one of four eccentricity levels (y = 0, 4.8, 9.6, 16.8), measured by degrees of visual angle (dva). For all models, the observer’s measurements xL, xR are assumed to follow a Gaussian distribution centered on the true stimulus yL, yR, respectively, with standard deviation σx(y) dependent on eccentricity level y. D: Possible eccentricity levels (in dva). E: Stimulus distribution for collinearity judgment task. When C = 1, the vertical position of the left line segment yL is drawn from a Gaussian distribution centered at y with fixed standard deviation σy; the vertical position of the right segment yR is then set equal to yL. When C = 0, yL and yR are independently drawn from the same Gaussian.

Cited by 0Open access