Search

In today’s automotive industry, digital technology trends such as Big Data, Digital Twin, and Hardware-in-the-loop simulations using synthetic data offer opportunities that have the potential to transform the entire industry towards being more software-oriented and thus more effective and environmentally friendly. In this paper, we propose generative models to synthesize car features related to vehicle speed: brake pressure, percentage of the pressed throttle pedal, engaged gear, and engine RPM. Synthetic data are essential to digitize Hardware-in-the-loop integration testing of the vehicle’s dashboard, navigation, or infotainment and for Digital Twin simulations. We trained models based on Multilayer Perceptron and bidirectional Long-Short Term Memory neural network for each feature. These models were evaluated on a real-world dataset and demonstrated sufficient accuracy in predicting the desired features. Combining our current research with previous work on generating a speed profile for an arbitrary trip, where Open Street Map data and elevation data are available, allows us to digitally drive this trip. At the time of writing, we are unaware of any similar data-driven approach for generating desired speed-related features.

A level generator is a tool that generates game levels from noise. Training a generator without a dataset suffers from feedback sparsity, since it is unlikely to generate a playable level via random exploration. A common solution is shaped rewards, which guides the generator to achieve subgoals towards level playability, but they consume effort to design and require game-specific domain knowledge. This paper proposes a novel approach to train generators without datasets or shaped rewards by learning at multiple level sizes starting from small sizes and up to the desired sizes. The denser feedback at small sizes negates the need for shaped rewards. Additionally, the generators learn to build levels at various sizes, including sizes they were not trained for. We apply our approach to train recurrent autoregressive generative flow networks (GFlowNets) for controllable level generation. We also adapt diversity sampling to be compatible with GFlowNets. The results show that our generators create diverse playable levels at various sizes for Sokoban, Zelda, and Danger Dave. When compared with controllable reinforcement learning level generators for Sokoban, the results show that our generators achieve better controllability and competitive diversity, while being 9× faster at training and level generation.

Abstract Deep learning has been successfully used for tasks in the 2D image domain. Research on 3D computer vision and deep geometry learning has also attracted attention. Considerable achievements have been made regarding feature extraction and discrimination of 3D shapes. Following recent advances in deep generative models such as generative adversarial networks, effective generation of 3D shapes has become an active research topic. Unlike 2D images with a regular grid structure, 3D shapes have various representations, such as voxels, point clouds, meshes, and implicit functions. For deep learning of 3D shapes, shape representation has to be taken into account as there is no unified representation that can cover all tasks well. Factors such as the representativeness of geometry and topology often largely affect the quality of the generated 3D shapes. In this survey, we comprehensively review works on deep-learning-based 3D shape generation by classifying and discussing them in terms of the underlying shape representation and the architecture of the shape generator. The advantages and disadvantages of each class are further analyzed. We also consider the 3D shape datasets commonly used for shape generation. Finally, we present several potential research directions that hopefully can inspire future works on this topic.

Normalizing flows have emerged as a powerful brand of generative models, as they not only allow for efficient sampling of complicated target distributions but also deliver density estimation by construction. We propose here an in-depth comparison of coupling and autoregressive flows, both based on symmetric (affine) and non-symmetric (rational quadratic spline) bijectors, considering four different architectures: real-valued non-Volume preserving (RealNVP), masked autoregressive flow (MAF), coupling rational quadratic spline (C-RQS), and autoregressive rational quadratic spline (A-RQS). We focus on a set of multimodal target distributions of increasing dimensionality ranging from 4 to 400. The performances were compared by means of different test statistics for two-sample tests, built from known distance measures: the sliced Wasserstein distance, the dimension-averaged one-dimensional Kolmogorov–Smirnov test, and the Frobenius norm of the difference between correlation matrices. Furthermore, we included estimations of the variance of both the metrics and the trained models. Our results indicate that the A-RQS algorithm stands out both in terms of accuracy and training speed. Nonetheless, all the algorithms are generally able, without too much fine-tuning, to learn complicated distributions with limited training data and in a reasonable time of the order of hours on a Tesla A40 GPU. The only exception is the C-RQS, which takes significantly longer to train, does not always provide good accuracy, and becomes unstable for large dimensionalities. All algorithms were implemented using TensorFlow2 and TensorFlow Probability and have been made available on GitHub.

Cognitive scientists believe that adaptable intelligent agents like humans perform spatial reasoning tasks by learned causal mental simulation. The problem of learning these simulations is called predictive world modeling. We present the first framework for a learning open-vocabulary predictive world model (OV-PWM) from sensor observations. The model is implemented through a hierarchical variational autoencoder (HVAE) capable of predicting diverse and accurate fully observed environments from accumulated partial observations. We show that the OV-PWM can model high-dimensional embedding maps of latent compositional embeddings representing sets of overlapping semantics inferable by sufficient similarity inference. The OV-PWM simplifies the prior two-stage closed-set PWM approach to the single-stage end-to-end learning method. CARLA simulator experiments show that the OV-PWM can learn compact latent representations and generate diverse and accurate worlds with fine details like road markings, achieving 69 mIoU over six query semantics on an urban evaluation sequence. We propose the OV-PWM as a versatile continual learning paradigm for providing spatio-semantic memory and learned internal simulation capabilities to future general-purpose mobile robots.

The auto-generation of urban roads can greatly improve efficiency and productivity in urban planning and designing. However, it has also raised concerns amongst researchers over the past decade. In this paper, we present an image-based urban road network generation framework using conditional diffusion models. We first trained a diffusion model capable of generating road images with similar characteristics to the ground truth using four context factors. Then, we used the trained model as the generator to synthesize road images conditioned in a geospatial context. Finally, we converted the generated road images into road networks with several post-processes. The experiments conducted in five cities of the United States showed that our model can generate reasonable road networks, maintaining the layouts and styles of real examples. Moreover, our model has the ability to show the obstructive effect of geographic barriers on urban roads. By comparing models with different context factors as input, we find that the model that considers all four factors generally performs the best. The most important factor in guiding the shape of road networks is intersections, implying that the development of urban roads is not only restricted by the natural environment but is more strongly influenced by human design.

Sequential recommender models should capture evolving user preferences over time, but there is a risk of obtaining biased results such as false positives and false negatives due to noisy interactions. Generative models effectively learn the underlying distribution and uncertainty of the given data to generate new data, and they exhibit robustness against noise. In particular, utilizing the Diffusion model, which generates data through a multi-step process of adding and removing noise, enables stable and effective recommendations. The Diffusion model typically leverages a Gaussian distribution with a mean fixed at zero, but there is potential for performance improvement in generative models by employing distributions with higher degrees of freedom. Therefore, we propose a Diffusion model-based sequential recommender model that uses a new noise distribution. The proposed model improves performance through a Weibull distribution with two parameters determining shape and scale, a modified Transformer architecture based on Macaron Net, normalized loss, and a learning rate warmup strategy. Experimental results on four types of real-world e-commerce data show that the proposed model achieved performance gains ranging from a minimum of 2.53% to a maximum of 13.52% across HR@K and NDCG@K metrics compared to the existing Diffusion model-based sequential recommender model.

Pattern recognition systems always misclassify anomalies, which can be dangerous for uninformed users. Therefore, anomalies must be filtered out from each classification. The main challenge for the anomaly filter design is the huge number of possible anomaly samples compared with the number of samples in the training set. Tailoring the filter for the given classifier is just the first step in this reduction. Paper tests the hypothesis that the filter trained in avoiding “near” anomalies will also refuse the “far” anomalies, and the anomaly detector is then just a classifier distinguishing between “far real” and “near anomaly” samples. As a “far real” samples generator was used, a Generative Adversarial Network (GAN) fake generator that transforms normally distributed random seeds into fakes similar to the training samples. The paper proves the assumption that seeds unused in fake training will generate anomalies. These seeds are distinguished according to their Chebyshev norms. While the fakes have seeds within the hypersphere with a given radius, the near anomalies have seeds within the sphere near cover. Experiments with various anomaly test sets have shown that GAN-based anomaly detectors create a reliable anti-anomaly shield using the abovementioned assumptions. The proposed anomaly detector is tailored to the given classifier, but its limitation is due to the need for the availability of the database on which the classifier was trained.

Recent advances in the field of artificial intelligence, specifically regarding deep learning methods, have invigorated research into novel ways for the exploration of chemical space. Compared to more traditional methods that rely on chemical fragments and combinatorial recombination deep generative models generate molecules in a non-transparent way that defies easy rationalization. However, this opaque nature also promises to explore uncharted chemical space in novel ways that do not rely on structural similarity directly. These aspects and the complexity of training such models makes model assessment regarding novelty, uniqueness, and distribution of generated molecules a central aspect. This perspective gives an overview of current methodologies for chemical space exploration with an emphasis on deep neural network approaches. Key aspects of generative models include choice of molecular representation, the targeted chemical space, and the methodology for assessing and validating chemical space coverage.

Predictions for physical systems often rely upon knowledge acquired from ensembles of entities, e.g. ensembles of cells in biological sciences. For qualitative and quantitative analysis, these ensembles are simulated with parametric families of mechanistic models (MMs). Two classes of methodologies, based on Bayesian inference and population of models, currently prevail in parameter estimation for physical systems. However, in Bayesian analysis, uninformative priors for MM parameters introduce undesirable bias. Here, we propose how to infer parameters within the framework of stochastic inverse problems (SIPs), also termed data-consistent inversion, wherein the prior targets only uncertainties that arise due to MM non-invertibility. To demonstrate, we introduce new methods to solve SIPs based on rejection sampling, Markov chain Monte Carlo, and generative adversarial networks (GANs). In addition, to overcome limitations of SIPs, we reformulate SIPs based on constrained optimization and present a novel GAN to solve the constrained optimization problem.

The predictive processing framework includes a broad set of ideas, which might be articulated and developed in a variety of ways, concerning how the brain may leverage predictive models when implementing perception, cognition, decision‐making, and motor control. This article provides an up‐to‐date introduction to the two most influential theories within this framework: predictive coding and active inference. The first half of the paper (Sections 2–5) reviews the evolution of predictive coding, from early ideas about efficient coding in the visual system to a more general model encompassing perception, cognition, and motor control. The theory is characterized in terms of the claims it makes at Marr's computational, algorithmic, and implementation levels of description, and the conceptual and mathematical connections between predictive coding, Bayesian inference, and variational free energy (a quantity jointly evaluating model accuracy and complexity) are explored. The second half of the paper (Sections 6–8) turns to recent theories of active inference. Like predictive coding, active inference models assume that perceptual and learning processes minimize variational free energy as a means of approximating Bayesian inference in a biologically plausible manner. However, these models focus primarily on planning and decision‐making processes that predictive coding models were not developed to address. Under active inference, an agent evaluates potential plans (action sequences) based on their expected free energy (a quantity that combines anticipated reward and information gain). The agent is assumed to represent the world as a partially observable Markov decision process with discrete time and discrete states. Current research applications of active inference models are described, including a range of simulation work, as well as studies fitting models to empirical data. The paper concludes by considering future research directions that will be important for further development of both models. [ABSTRACT FROM AUTHOR]

Despite recent breakthroughs in the domain of implicit generative models, the task of evaluating these models remains a challenging task. With no single metric to assess overall performance, various existing metrics only offer partial information. This issue is further compounded for unintuitive data types such as time series, where manual inspection is infeasible. This deficiency hinders the confident application of modern implicit generative models on time series data. To alleviate this problem, we propose two new metrics, the InceptionTime Score (ITS) and the Fréchet InceptionTime Distance (FITD), to assess the quality of class-conditional generative models on time series data. We conduct extensive experiments on 80 different datasets to study the discriminative capabilities of proposed metrics alongside two existing evaluation metrics: Train on Synthetic Test on Real (TSTR) and Train on Real Test on Synthetic (TRTS). Our evaluations reveal that the proposed assessment evaluation metrics, i.e., ITS and FITD in combination with TSTR, can accurately assess class-conditional generative model performance and detect common issues in implicit generative models. Our findings suggest that the proposed evaluation framework can be a valuable tool for confidently applying modern implicit generative models in time series analysis. [ABSTRACT FROM AUTHOR]

Until recently, conventional biochemical staining had the undisputed status as well-established benchmark for most biomedical problems related to clinical diagnostics, fundamental research and biotechnology. Despite this role as gold-standard, staining protocols face several challenges, such as a need for extensive, manual processing of samples, substantial time delays, altered tissue homeostasis, limited choice of contrast agents, 2D imaging instead of 3D tomography and many more. Label-free optical technologies, on the other hand, do not rely on exogenous and artificial markers, by exploiting intrinsic optical contrast mechanisms, where the specificity is typically less obvious to the human observer. Over the past few years, digital staining has emerged as a promising concept to use modern deep learning for the translation from optical contrast to established biochemical contrast of actual stainings. In this review article, we provide an in-depth analysis of the current state-of-the-art in this field, suggest methods of good practice, identify pitfalls and challenges and postulate promising advances towards potential future implementations and applications. [ABSTRACT FROM AUTHOR]

Cardiovascular diseases (CVDs) are a major global health concern, causing significant morbidity and mortality. AI's integration with healthcare offers promising solutions, with data-driven techniques, including ECG analysis, emerging as powerful tools. However, privacy concerns pose a major barrier to distributing healthcare data for addressing data-driven CVD classification. To address confidentiality issues related to sensitive health data distribution, we propose leveraging artificially synthesized data generation. Our contribution introduces a novel diffusion-based model coupled with a State Space Augmented Transformer. This synthesizes conditional 12-lead electrocardiograms based on the 12 multilabeled heart rhythm classes of the PTB-XL dataset, with each lead depicting the heart's electrical activity from different viewpoints. Recent advances establish diffusion models as groundbreaking generative tools, while the State Space Augmented Transformer captures long-term dependencies in time series data. The quality of generated samples was assessed using metrics like Dynamic Time Warping (DTW) and Maximum Mean Discrepancy (MMD). To evaluate authenticity, we assessed the similarity of performance of a pre-trained classifier on both generated and real ECG samples. [ABSTRACT FROM AUTHOR]

Copyright of Romanian Journal of Intellectual Property Law / Revista Română de Dreptul Proprietăţii Intelectuale is the property of Universul Juridic Publishing House and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Recent studies have shown that modern deep neural network classifiers are easy to fool, assuming that an adversary is able to slightly modify their inputs. Many papers have proposed adversarial attacks, defenses and methods to measure robustness to such adversarial perturbations. However, most commonly considered adversarial examples are based on perturbations in the input space of the neural network that are unlikely to arise naturally. Recently, especially in computer vision, researchers discovered "natural" perturbations, such as rotations, changes of brightness, or more high-level changes, but these perturbations have not yet been systematically used to measure the performance of classifiers. In this paper, we propose several metrics to measure robustness of classifiers to natural adversarial examples, and methods to evaluate them. These metrics, called latent space performance metrics, are based on the ability of generative models to capture probability distributions. On four image classification case studies, we evaluate the proposed metrics for several classifiers, including ones trained in conventional and robust ways. We find that the latent counterparts of adversarial robustness are associated with the accuracy of the classifier rather than its conventional adversarial robustness, but the latter is still reflected on the properties of found latent perturbations. In addition, our novel method of finding latent adversarial perturbations demonstrates that these perturbations are often perceptually small. [ABSTRACT FROM AUTHOR]

Generative adversarial networks have succeeded promising results in the medical imaging field. One of the most significant challenges in this regard is the lack of or limited data sharing. In our work, an approach for combining generative adversarial network (GAN) and variational autoencoder (VAE) models has been proposed to improve the accuracy and efficiency of medical image analysis tasks. Our approach leverages the capacity of VAEs to acquire condensed feature representations, and the ability of GANs to generate high-quality synthetic images to learn an embedding that keeps high-level abstract visual qualities. Inception score (IS) and Fréchet inception distance (FID) score have been generated in order to demonstrate the high quality of images. Based on the score results, our approach demonstrates the potential of VAE-GAN fusion models and clearly outperforms existing methods on a variety of medical image analysis tasks. The suggested algorithm is explained, as are the results and evaluations. [ABSTRACT FROM AUTHOR]

Understanding the mechanism by which cells coordinate their differentiation and migration is critical to our understanding of many fundamental processes such as wound healing, disease progression, and developmental biology. Mathematical models have been an essential tool for testing and developing our understanding, such as models of cells as soft spherical particles, reaction-diffusion systems that couple cell movement to environmental factors, and multi-scale multi-physics simulations that combine bottom-up rule-based models with continuum laws. However, mathematical models can often be loosely related to data or have so many parameters that model behaviour is weakly constrained. Recent methods in machine learning introduce new means by which models can be derived and deployed. In this review, we discuss examples of mathematical models of aspects of developmental biology, such as cell migration, and how these models can be combined with these recent machine learning methods. [ABSTRACT FROM AUTHOR]

Generative modeling and representation learning of tandem mass spectrometry data aim to learn an interpretable and instrument-agnostic digital representation of metabolites directly from MS/MS spectra. Interpretable and instrument-agnostic digital representations would facilitate comparisons of MS/MS spectra between instrument vendors and enable better and more accurate queries of large MS/MS spectra databases for metabolite identification. In this study, we apply generative modeling and representation learning using variational autoencoders to understand the extent to which tandem mass spectra can be disentangled into their factors of generation (e.g., collision energy, ionization mode, instrument type, etc.) with minimal prior knowledge of the factors. We find that variational autoencoders can disentangle tandem mass spectra data with the proper choice of hyperparameters into meaningful latent representations aligned with known factors of variation. We develop a two-step approach to facilitate the selection of models that are disentangled, which could be applied to other complex and high-dimensional data sets. [ABSTRACT FROM AUTHOR]

Accurately identifying and contouring the organs at risk (OARs) is a crucial step in radiation treatment planning for precise dose calculation. This task becomes especially challenging in computed tomography (CT) images due to the irregular boundaries of the organs under study. The method currently employed in clinical practice is the manual contouring of CT images, which tends to be highly tedious and time‐consuming. The results are also prone to variations depending on the observer's skill level, environment, or equipment types. A deep learning‐based automatic contouring technique for segmenting OARs would help eliminate these problems and generate consistent results with minimal time and human effort. Our approach is to design a conditional generative adversarial network (GAN)‐based technique for the semantic segmentation of OARs in abdominal CT images. The residual blocks of the generator network have a multi‐scale context layer that explores more generic characteristics, greatly enhancing performance and lowering losses. A comparative analysis is undertaken based on various assessment measures widely employed in segmentation. The results show substantial improvement, with mean dice scores of 98.0%, 96.6%, 98.2%, and 86.1% for the respective organs—liver, kidney, spleen, and pancreas—in the abdominal CT. The proposed GAN‐based model could accurately segment the four abdominal organs, including the liver, kidney, spleen, and pancreas. The obtained results prove that the suggested model is able to compete with existing state‐of‐the‐art abdominal OAR segmentation techniques. [ABSTRACT FROM AUTHOR]

Deep learning has been successfully used for tasks in the 2D image domain. Research on 3D computer vision and deep geometry learning has also attracted attention. Considerable achievements have been made regarding feature extraction and discrimination of 3D shapes. Following recent advances in deep generative models such as generative adversarial networks, effective generation of 3D shapes has become an active research topic. Unlike 2D images with a regular grid structure, 3D shapes have various representations, such as voxels, point clouds, meshes, and implicit functions. For deep learning of 3D shapes, shape representation has to be taken into account as there is no unified representation that can cover all tasks well. Factors such as the representativeness of geometry and topology often largely affect the quality of the generated 3D shapes. In this survey, we comprehensively review works on deep-learning-based 3D shape generation by classifying and discussing them in terms of the underlying shape representation and the architecture of the shape generator. The advantages and disadvantages of each class are further analyzed. We also consider the 3D shape datasets commonly used for shape generation. Finally, we present several potential research directions that hopefully can inspire future works on this topic. [ABSTRACT FROM AUTHOR]

The rise of explainable natural language processing spurred a bulk of work on datasets augmented with human explanations, as well as technical approaches to leverage them. Notably, generative large language models offer new possibilities, as they can output a prediction as well as an explanation in natural language. This work investigates the capabilities of fine-tuned text-to-text transfer Transformer (T5) models for commonsense reasoning and explanation generation. Our experiments suggest that while self-rationalizing models achieve interesting results, a significant gap remains: classifiers consistently outperformed self-rationalizing models, and a substantial fraction of model-generated explanations are not valid. Furthermore, training with expressive free-text explanations substantially altered the inner representation of the model, suggesting that they supplied additional information and may bridge the knowledge gap. Our code is publicly available, and the experiments were run on open-access datasets, hence allowing full reproducibility. [ABSTRACT FROM AUTHOR]

In today's automotive industry, digital technology trends such as Big Data, Digital Twin, and Hardware-in-the-loop simulations using synthetic data offer opportunities that have the potential to transform the entire industry towards being more software-oriented and thus more effective and environmentally friendly. In this paper, we propose generative models to synthesize car features related to vehicle speed: brake pressure, percentage of the pressed throttle pedal, engaged gear, and engine RPM. Synthetic data are essential to digitize Hardware-in-the-loop integration testing of the vehicle's dashboard, navigation, or infotainment and for Digital Twin simulations. We trained models based on Multilayer Perceptron and bidirectional Long-Short Term Memory neural network for each feature. These models were evaluated on a real-world dataset and demonstrated sufficient accuracy in predicting the desired features. Combining our current research with previous work on generating a speed profile for an arbitrary trip, where Open Street Map data and elevation data are available, allows us to digitally drive this trip. At the time of writing, we are unaware of any similar data-driven approach for generating desired speed-related features. [ABSTRACT FROM AUTHOR]

Early adversity can change educational, cognitive, and mental health outcomes. However, the neural processes through which early adversity exerts these effects remain largely unknown. We used generative network modeling of the mouse connectome to test whether unpredictable postnatal stress shifts the constraints that govern the organization of the structural connectome. A model that trades off the wiring cost of long‐distance connections with topological homophily (i.e., links between regions with shared neighbors) generated simulations that successfully replicate the rodent connectome. The imposition of early life adversity shifted the best‐performing parameter combinations toward zero, heightening the stochastic nature of the generative process. Put simply, unpredictable postnatal stress changes the economic constraints that reproduce rodent connectome organization, introducing greater randomness into the development of the simulations. While this change may constrain the development of cognitive abilities, it could also reflect an adaptive mechanism that facilitates effective responses to future challenges. [ABSTRACT FROM AUTHOR]

Generative Adversarial Networks (GANs) are a type of deep learning architecture that uses two networks namely a generator and a discriminator that, by competing against each other, pursue to create realistic but previously unseen samples. They have become a popular research topic in recent years, particularly for image processing and synthesis, leading to many advances and applications in various fields. With the profusion of published works and interest from professionals of different areas, surveys on GANs are necessary, mainly for those who aim starting on this topic. In this work, we cover the basics and notable architectures of GANs, focusing on their applications in image generation. We also discuss how the challenges to be addressed in GANs architectures have been faced, such as mode coverage, stability, convergence, and evaluating image quality using metrics. [Display omitted] • A review on GANs for image generation, aiming at readers who are new to the area. • A comprehensive overview of GAN fundamentals, and methods to address the most common issues. • A detailed explanation of how various works applied GANs in image-based applications. • A discussion of future directions for this area. [ABSTRACT FROM AUTHOR]

• We propose a disentangled representation that models the labels and their features. • Our proposal includes a novel dependency structure and a two-step learning procedure. • It allows accurate control of labels and their features in the generated images. • We show several examples of label and feature manipulation for 2D images. • Our approach improves disentanglement properties and the quality of generated images. Two recent works have shown the benefit of modeling both high-level factors and their related features to learn disentangled representations with variational autoencoders (VAE). We propose here a novel VAE-based approach that follows this principle. Inspired by conditional VAE, the features are no longer treated as random variables over which integration must be performed. Instead, they are deterministically computed from the input data using a neural network whose parameters can be estimated jointly with those of the decoder and of the encoder. Moreover, the quality of the generated images has been improved by using discrete latent variables and a two-step learning procedure, which makes it possible to increase the size of the latent space without altering the disentanglement properties of the model. Results obtained on two different datasets validate the proposed approach that achieves better performance than the two aforementioned works in terms of disentanglement, while providing higher quality images. [ABSTRACT FROM AUTHOR]

Diffusion models have experienced a surge of interest as highly expressive yet efficiently trainable probabilistic models. We show that these models are an excellent fit for synthesising human motion that co-occurs with audio, e.g., dancing and co-speech gesticulation, since motion is complex and highly ambiguous given audio, calling for a probabilistic description. Specifically, we adapt the DiffWave architecture to model 3D pose sequences, putting Conformers in place of dilated convolutions for improved modelling power. We also demonstrate control over motion style, using classifier-free guidance to adjust the strength of the stylistic expression. Experiments on gesture and dance generation confirm that the proposed method achieves top-of-the-line motion quality, with distinctive styles whose expression can be made more or less pronounced. We also synthesise path-driven locomotion using the same model architecture. Finally, we generalise the guidance procedure to obtain product-of-expert ensembles of diffusion models and demonstrate how these may be used for, e.g., style interpolation, a contribution we believe is of independent interest. [ABSTRACT FROM AUTHOR]

We introduce 3DShape2VecSet, a novel shape representation for neural fields designed for generative diffusion models. Our shape representation can encode 3D shapes given as surface models or point clouds, and represents them as neural fields. The concept of neural fields has previously been combined with a global latent vector, a regular grid of latent vectors, or an irregular grid of latent vectors. Our new representation encodes neural fields on top of a set of vectors. We draw from multiple concepts, such as the radial basis function representation, and the cross attention and self-attention function, to design a learnable representation that is especially suitable for processing with transformers. Our results show improved performance in 3D shape encoding and 3D shape generative modeling tasks. We demonstrate a wide variety of generative applications: unconditioned generation, category-conditioned generation, text-conditioned generation, point-cloud completion, and image-conditioned generation. Code: https://1zb.github.io/3DShape2VecSet/. [ABSTRACT FROM AUTHOR]

This work considers gradient-based mesh optimization, where we iteratively optimize for a 3D surface mesh by representing it as the isosurface of a scalar field, an increasingly common paradigm in applications including photogrammetry, generative modeling, and inverse physics. Existing implementations adapt classic isosurface extraction algorithms like Marching Cubes or Dual Contouring; these techniques were designed to extract meshes from fixed, known fields, and in the optimization setting they lack the degrees of freedom to represent high-quality feature-preserving meshes, or suffer from numerical instabilities. We introduce FlexiCubes, an isosurface representation specifically designed for optimizing an unknown mesh with respect to geometric, visual, or even physical objectives. Our main insight is to introduce additional carefully-chosen parameters into the representation, which allow local flexible adjustments to the extracted mesh geometry and connectivity. These parameters are updated along with the underlying scalar field via automatic differentiation when optimizing for a downstream task. We base our extraction scheme on Dual Marching Cubes for improved topological properties, and present extensions to optionally generate tetrahedral and hierarchically-adaptive meshes. Extensive experiments validate FlexiCubes on both synthetic benchmarks and real-world applications, showing that it offers significant improvements in mesh quality and geometric fidelity. [ABSTRACT FROM AUTHOR]

While Machine Learning has become the holy grail of modern-day computing, it has many security flaws that have yet to be addressed and resolved. Adversarial attacks are one of these security flaws, in which an attacker appends noise to data samples that machine learning models take as input with the aim of fooling the model. Various adversarial training methods have been proposed that augment adversarial examples in the training dataset for defense against such attacks. However, a general limitation exists where a robust model can only protect itself against adversarial attacks that are known or similar to those it was trained on. To address this limitation, this paper proposes a Universal Adversarial Training algorithm using adversarial examples generated by an Auxiliary Classifier Generative Adversarial Network (AC-GAN) in parallel with other data augmentation techniques, such as the mixup method. This method builds on a previously proposed technique, Adversarial Training, in which adversarial examples produced by gradient-based methods are augmented and added to the training data. Our method improves the AC-GAN architecture for adversarial example generation to make it more suitable for adversarial training by updating different loss terms and testing its performance against various attacks compared to other robust adversarial models. In this way, it becomes apparent that generative models are better suited for boosting adversarial robustness through adversarial training. When tested using various attack types, our proposed model had an average accuracy of 97.48% on the MNIST dataset and 94.02% on the CelebA dataset, proving that generative models have a higher chance of boosting adversarial security through adversarial training. [ABSTRACT FROM AUTHOR]

Abstract The flourishment of machine learning and deep learning methods has boosted the development of cheminformatics, especially regarding the application of drug discovery and new material exploration. Lower time and space expenses make it possible for scientists to search the enormous chemical space. Recently, some work combined reinforcement learning strategies with recurrent neural network (RNN)-based models to optimize the property of generated small molecules, which notably improved a batch of critical factors for these candidates. However, a common problem among these RNN-based methods is that several generated molecules have difficulty in synthesizing despite owning higher desired properties such as binding affinity. However, RNN-based framework better reproduces the molecule distribution among the training set than other categories of models during molecule exploration tasks. Thus, to optimize the whole exploration process and make it contribute to the optimization of specified molecules, we devised a light-weighted pipeline called Magicmol; this pipeline has a re-mastered RNN network and utilize SELFIES presentation instead of SMILES. Our backbone model achieved extraordinary performance while reducing the training cost; moreover, we devised reward truncate strategies to eliminate the model collapse problem. Additionally, adopting SELFIES presentation made it possible to combine STONED-SELFIES as a post-processing procedure for specified molecule optimization and quick chemical space exploration.

Abstract Artificial intelligence (AI)-based molecular design methods, especially deep generative models for generating novel molecule structures, have gratified our imagination to explore unknown chemical space without relying on brute-force exploration. However, whether designed by AI or human experts, the molecules need to be accessibly synthesized and biologically evaluated, and the trial-and-error process remains a resources-intensive endeavor. Therefore, AI-based drug design methods face a major challenge of how to prioritize the molecular structures with potential for subsequent drug development. This study indicates that common filtering approaches based on traditional screening metrics fail to differentiate AI-designed molecules. To address this issue, we propose a novel molecular filtering method, MolFilterGAN, based on a progressively augmented generative adversarial network. Comparative analysis shows that MolFilterGAN outperforms conventional screening approaches based on drug-likeness or synthetic ability metrics. Retrospective analysis of AI-designed discoidin domain receptor 1 (DDR1) inhibitors shows that MolFilterGAN significantly increases the efficiency of molecular triaging. Further evaluation of MolFilterGAN on eight external ligand sets suggests that MolFilterGAN is useful in triaging or enriching bioactive compounds across a wide range of target types. These results highlighted the importance of MolFilterGAN in evaluating molecules integrally and further accelerating molecular discovery especially combined with advanced AI generative models.