Link to paper The full paper is available here. You can also find the paper on PapersWithCode here. Abstract Novel optimization algorithm called DroNeRF for autonomous positioning of monocular camera drones Used to generate new views of an object or scene from a set of input images Calculates optimized pose for individual drones without using external localization system Evaluates quality of generated novel views using perceptual metrics like PSNR and SSIM Paper Content Iii....

Link to paper The full paper is available here. You can also find the paper on PapersWithCode here. Abstract Denoising Diffusion models are proficient for generative sampling. Techniques such as BTD have been proposed to reduce the number of network calls. TRACT is a new method that extends BTD. TRACT improves FID by up to 2.4x on the same architecture. TRACT achieves new single-step Denoising Diffusion Implicit Models state-of-the-art FID. PyTorch implementation will be released soon....

Link to paper The full paper is available here. You can also find the paper on PapersWithCode here. Abstract Transformers have been successful across many domains, but their learning mechanics are not well understood. Recent research has begun to study the representational aspects of transformers, but there is no guarantee the learning dynamics will converge. This paper provides a mechanistic understanding of how transformers learn “semantic structure” by combining experiments on synthetic data, Wikipedia data, and mathematical analysis....

Link to paper The full paper is available here. You can also find the paper on PapersWithCode here. Abstract Category theory has been applied to various scientific domains. DisCoPyro combines categorical structures with amortized variational inference. DisCoPyro can be applied to program learning for variational autoencoders. DisCoPyro provides mathematical foundations and concrete applications. DisCoPyro is compared to other models (e.g. neuro-symbolic models). DisCoPyro could contribute to the development of artificial general intelligence....

Link to paper The full paper is available here. You can also find the paper on PapersWithCode here. Abstract Semantic occupancy perception is essential for autonomous driving. Existing benchmarks lack diversity in urban scenes and only evaluate front-view predictions. OpenOccupancy is the first surrounding semantic occupancy perception benchmark. Annotations rely on LiDAR points superimposition, which can miss some occupancy labels. Augmenting And Purifying (AAP) pipeline is used to densify the annotations....

Link to paper The full paper is available here. You can also find the paper on PapersWithCode here. Abstract Need for large-scale high-quality text datasets BigScience workshop formed to research and train large language models ROOTS corpus created, 1.6TB dataset spanning 59 languages BLOOM language model trained using ROOTS corpus Large initial subset of corpus released with processing tools Paper Content Introduction BigScience1 is a one-year open collaborative research initiative Goal was to train an open-access, massively multilingual language model Engaged in ethical, sociopolitical, and data governance issues Four working groups: Data Governance, Data Sourcing and Preparation, Privacy, Legal Scholarship Released a large subset of ROOTS Released data tools used to curate, source, clean and inspect constituent datasets Outline of the paper Collected a web-scale dataset covering 59 languages 46 natural languages and 13 programming languages 62% of text from community-selected and documented list of language data sources 38% of text from pre-processed web crawl, OSCAR Filtered with help of native speakers Related work Pre-trained models are used in natural language processing Performance is based on model size and dataset size/quality Recent models trained on up to 1....

Link to paper The full paper is available here. You can also find the paper on PapersWithCode here. Abstract Optimal transport-based distributionally robust optimization problems studied Fictitious adversary (nature) can choose distribution of uncertain problem parameters Robustification related to various forms of variation and Lipschitz regularization Conditions for existence and computability of Nash equilibrium between decision-maker and nature Paper Content Introduction Stochastic optimization methods are becoming popular in machine learning. A stochastic optimization problem aims to minimize the expected value of an uncertainty-affected loss function....

Link to paper The full paper is available here. You can also find the paper on PapersWithCode here. Abstract Markov state models are used to interpret molecular dynamics trajectories. Structurally distinct conformations are needed to understand the biomolecular process. Dihedral angles and interresidue distances are used as input coordinates. Contacts are used to define and select contact distances. Low-pass filtering and correlation-based characterization of states are used. States of the Markov model are discriminated by the features....

Link to paper The full paper is available here. You can also find the paper on PapersWithCode here. Abstract Focuses on a novel optimization problem where the objective function can only be evaluated through a ranking oracle. RLHF is an example of this problem, used to improve the quality of LLMs with human guidance. Proposes ZO-RankSGD, a zeroth-order optimization algorithm with a theoretical guarantee. Can be applied to policy search problems in reinforcement learning with only a ranking oracle of the episode reward....

Link to paper The full paper is available here. You can also find the paper on PapersWithCode here. Abstract Many works have focused on reducing the number of FLOPs to design fast neural networks. Reducing FLOPs does not necessarily lead to a similar level of reduction in latency. Low FLOPS is mainly due to frequent memory access of operators, especially the depthwise convolution. Proposed a novel partial convolution (PConv) to extract spatial features more efficiently....