搜索结果：Mobile DNA

Rapid advancements in large language models (LLMs) have increased interest in deploying them on mobile devices for on-device AI applications. Mobile users interact differently with LLMs compared to desktop users, creating unique expectations and data biases. Current benchmark datasets primarily target at server and desktop environments, and there is a notable lack of extensive datasets specifically designed for mobile contexts. Additionally, mobile devices face strict limitations in storage and computing resources, constraining model size and capabilities, thus requiring optimized efficiency and prioritized knowledge. To address these challenges, we introduce Mobile-MMLU, a large-scale benchmark dataset tailored for mobile intelligence. It consists of 16,186 questions across 80 mobile-related fields, designed to evaluate LLM performance in realistic mobile scenarios. A challenging subset, Mobile-MMLU-Pro, provides advanced evaluation similar in size to MMLU-Pro but significantly more difficult than our standard full set. Both benchmarks use multiple-choice, order-invariant questions focused on practical mobile interactions, such as recipe suggestions, travel planning, and essential

Unified multimodal models can both understand and generate visual content within a single architecture. Existing models, however, remain data-hungry and too heavy for deployment on edge devices. We present Mobile-O, a compact vision-language-diffusion model that brings unified multimodal intelligence to a mobile device. Its core module, the Mobile Conditioning Projector (MCP), fuses vision-language features with a diffusion generator using depthwise-separable convolutions and layerwise alignment. This design enables efficient cross-modal conditioning with minimal computational cost. Trained on only a few million samples and post-trained in a novel quadruplet format (generation prompt, image, question, answer), Mobile-O jointly enhances both visual understanding and generation capabilities. Despite its efficiency, Mobile-O attains competitive or superior performance compared to other unified models, achieving 74% on GenEval and outperforming Show-O and JanusFlow by 5% and 11%, while running 6x and 11x faster, respectively. For visual understanding, Mobile-O surpasses them by 15.3% and 5.1% averaged across seven benchmarks. Running in only ~3s per 512x512 image on an iPhone, Mobile-O

3D Gaussian Splatting (3DGS) has emerged as a powerful representation for high-quality rendering across a wide range of applications.However, its high computational demands and large storage costs pose significant challenges for deployment on mobile devices. In this work, we propose a mobile-tailored real-time Gaussian Splatting method, dubbed Mobile-GS, enabling efficient inference of Gaussian Splatting on edge devices. Specifically, we first identify alpha blending as the primary computational bottleneck, since it relies on the time-consuming Gaussian depth sorting process. To solve this issue, we propose a depth-aware order-independent rendering scheme that eliminates the need for sorting, thereby substantially accelerating rendering. Although this order-independent rendering improves rendering speed, it may introduce transparency artifacts in regions with overlapping geometry due to the scarcity of rendering order. To address this problem, we propose a neural view-dependent enhancement strategy, enabling more accurate modeling of view-dependent effects conditioned on viewing direction, 3D Gaussian geometry, and appearance attributes. In this way, Mobile-GS can achieve both high

Mobile video applications today have attracted significant attention. Deep learning model (e.g. deep neural network, DNN) compression is widely used to enable on-device inference for facilitating robust and private mobile video applications. The compressed DNN, however, is vulnerable to the agnostic data drift of the live video captured from the dynamically changing mobile scenarios. To combat the data drift, mobile ends rely on edge servers to continuously evolve and re-compress the DNN with freshly collected data. We design a framework, AdaEvo, that efficiently supports the resource-limited edge server handling mobile DNN evolution tasks from multiple mobile ends. The key goal of AdaEvo is to maximize the average quality of experience (QoE), e.g. the proportion of high-quality DNN service time to the entire life cycle, for all mobile ends. Specifically, it estimates the DNN accuracy drops at the mobile end without labels and performs a dedicated video frame sampling strategy to control the size of retraining data. In addition, it balances the limited computing and memory resources on the edge server and the competition between asynchronous tasks initiated by different mobile user

The proton-magnetic reaction is commonly used in MRI machines with a strong magnetic field of over 1 T, while this study hypothesized that the electron magnetic reaction of hydrogen could affect the hydrogen bonds of double-stranded DNA (dsDNA) at a low magnetic field below 0.01 T. The goal is to develop a hydrogen bonding magnetic reaction-based gene regulation (HBMR-GR) system. The polarities of DNA base pairs are derived from the relative electrostatic charge between purines and pyrimidines, which become positively and negatively charged, respectively. The Pyu dsDNAs with pyrimidine(s)-purine(s) sequences, ds3T3A, ds3C3G, and ds3C3A, showed stronger DNA hybridization potential, increased infrared absorption at 3400-3200 cm-1, and a unique DNA conformation in HPLC analysis compared to the corresponding Puy dsDNAs. To target the three-dimensional structure of dsDNA based on the DNA base pair polarities, one can use cyclic electromagnetic DNA simulation (CEDS) with approximately 25% efficiency for randomly oriented dsDNAs. CEDS was found to induce sequence-specific hybridization of target oligo-dsDNAs in 0.005M NaCl solution and sequence-specific conformation of oligo-dsDNAs in 0.1

This paper introduces a novel framework for DNA sequence generation, comprising two key components: DiscDiff, a Latent Diffusion Model (LDM) tailored for generating discrete DNA sequences, and Absorb-Escape, a post-training algorithm designed to refine these sequences. Absorb-Escape enhances the realism of the generated sequences by correcting `round errors' inherent in the conversion process between latent and input spaces. Our approach not only sets new standards in DNA sequence generation but also demonstrates superior performance over existing diffusion models, in generating both short and long DNA sequences. Additionally, we introduce EPD-GenDNA, the first comprehensive, multi-species dataset for DNA generation, encompassing 160,000 unique sequences from 15 species. We hope this study will advance the generative modelling of DNA, with potential implications for gene therapy and protein production.

We introduce $\textit{scadnano}$ (https://scadnano.org) (short for "scriptable cadnano"), a computational tool for designing synthetic DNA structures. Its design is based heavily on cadnano, the most widely-used software for designing DNA origami, with three main differences: 1. scadnano runs entirely in the browser, with $\textit{no software installation}$ required. 2. scadnano designs, while they can be edited manually, can also be created and edited by a $\textit{well-documented Python scripting library}$, to help automate tedious tasks. 3. The scadnano file format is $\textit{easily human-readable}$. This goal is closely aligned with the scripting library, intended to be helpful when debugging scripts or interfacing with other software. The format is also somewhat more expressive than that of cadnano, able to describe a broader range of DNA structures than just DNA origami.

The idea of computing with DNA was given by Tom Head in 1987, however in 1994 in a seminal paper, the actual successful experiment for DNA computing was performed by Adleman. The heart of the DNA computing is the DNA hybridization, however, it is also the source of errors. Thus the success of the DNA computing depends on the error control techniques. The classical coding theory techniques have provided foundation for the current information and communication technology (ICT). Thus it is natural to expect that coding theory will be the foundational subject for the DNA computing paradigm. For the successful experiments with DNA computing usually we design DNA strings which are sufficiently dissimilar. This leads to the construction of a large set of DNA strings which satisfy certain combinatorial and thermodynamic constraints. Over the last 16 years, many approaches such as combinatorial, algebraic, computational have been used to construct such DNA strings. In this work, we survey this interesting area of DNA coding theory by providing key ideas of the area and current known results.

We study attention in mobile Augmented Reality (AR) using object recall as a proxy outcome. We observe that the ability to recall an object (physical or virtual) that was encountered in a mobile AR experience depends on many possible impact factors and attributes, with some objects being readily recalled while others are not, and some people recalling objects overall much better or worse than others. This opens up a potential cognitive attack in which adversaries might create conditions that make an AR user not recall certain potentially mission-critical objects. We explore whether a calibrated predictor of object recall can help shield against such cognitive attacks. We pool data from four mobile AR studies (with a total of 1,152 object recall probes) and fit a Partial Least Squares Structural Equation Model (PLS-SEM) with formative Object, Scene, and User State composites predicting recall, also benchmarking against Random Forest and multilayer perceptron classifiers. PLS-SEM attains the best F1 score in three of four studies. Additionally, path estimates identify lighting, augmentation density, AR registration stability, cognitive load, and AR familiarity as primary drivers. The

The characterization of the long-range order and fractal properties of DNA sequences has proved a difficult though highly rewarding task due mainly to the mosaic character of DNA consisting of many interwoven patches of various lengths with different nucleotide constitutions. We apply here a recently proposed generalization of the detrended fluctuation analysis method to show that the DNA walk construction, in which the DNA sequence is viewed as a time series, exhibits a monofractal structure regardless of the existence of local trends in the series. In addition, we point out that the monofractal structure of the DNA walks carries over to an apparently alternative graphical construction given by the projection of the DNA walk into the $d$ spatial coordinates, termed DNA trails. In particular, we calculate the fractal dimension $D_t$ of the DNA trails using a well-known result of fractal theory linking $D_t$ to the Hurst exponent $H$ of the corresponding DNA walk. Comparison with estimates obtained by the standard box-counting method allows the evaluation of both finite-length and local trends effects.

The ARMrayan Multimedia Mobile CMS (Content Management System) is the first mobile CMS that gives the opportunity to users for creating multimedia J2ME mobile applications with their desired content, design and logo; simply, without any need for writing even a line of code. The low-level programming and compatibility problems of the J2ME, along with UI designing difficulties, makes it hard for most people -even programmers- to broadcast their content to the widespread mobile phones used by nearly all people. This system provides user-friendly, PC-based tools for creating a tree index of pages and inserting multiple multimedia contents (e.g. sound, video and picture) in each page for creating a J2ME mobile application. The output is a stand-alone Java mobile application that has a user interface, shows texts and pictures and plays music and videos regardless of the type of devices used as long as the devices support the J2ME platform. Bitmap fonts have also been used thus Middle Eastern languages can be easily supported on all mobile phone devices. We omitted programming concepts for users in order to simplify multimedia content-oriented mobile application designing for use in educa

Emerging Artificial Intelligence of Things (AIoT) applications desire online prediction using deep neural network (DNN) models on mobile devices. However, due to the movement of devices, unfamiliar test samples constantly appear, significantly affecting the prediction accuracy of a pre-trained DNN. In addition, unstable network connection calls for local model inference. In this paper, we propose a light-weight scheme, called Anole, to cope with the local DNN model inference on mobile devices. The core idea of Anole is to first establish an army of compact DNN models, and then adaptively select the model fitting the current test sample best for online inference. The key is to automatically identify model-friendly scenes for training scene-specific DNN models. To this end, we design a weakly-supervised scene representation learning algorithm by combining both human heuristics and feature similarity in separating scenes. Moreover, we further train a model classifier to predict the best-fit scene-specific DNN model for each test sample. We implement Anole on different types of mobile devices and conduct extensive trace-driven and real-world experiments based on unmanned aerial vehicle

In this paper, we present the notion of "mobile 3C systems in which the "Communications", "Computing", and "Caching" (i.e., 3C) make up the three primary resources/funcationalties, akin to the three primary colors, for a mobile system. We argue that in future mobile networks, the roles of computing and caching are as intrinsic and essential as communications, and only the collective usage of these three primary resources can support the sustainable growth of mobile systems. By defining the 3C resources in their canonical forms, we reveal the important fact that "caching" affects the mobile system performance by introducing non-causality into the system, whereas "computing" achieves capacity gains by performing logical operations across mobile system entities. Many existing capacity-enhancing techniques such as coded multicast, collaborative transmissions, and proactive content pushing can be cast into the native 3C framework for analytical tractability. We further illustrate the mobile 3C concepts with practical examples, including a system on broadcast-unicast convergence for massive media content delivery. The mobile 3C design paradigm opens up new possibilities as well as key re

We scrutinize the effect of polyvalent ions on polymer-DNA interactions. We extend a recently developed test charge theory to the case of a stiff polymer interacting with a DNA molecule in an electrolyte mixture. The theory accounts for one-loop level electrostatic correlation effects such as the ionic cloud deformation around the strongly charged DNA molecule as well as image-charge forces induced by the low DNA permittivity. Our model can reproduce and explain various characteristics of the experimental phase diagrams for polymer solutions. First, the addition of polyvalent cations to the electrolyte solution results in the attraction of the negatively charged polymer by the DNA molecule. The glue of the like-charge attraction is the enhanced shielding of the polymer charges by the dense counterion layer at the DNA surface. Secondly, through the shielding of the DNA-induced electrostatic potential, mono- and polyvalent cations of large concentration both suppress the like-charge attraction. Within the same formalism, we also predict a new opposite-charge repulsion effect between the DNA molecule and a positively charged polymer. In the presence of polyvalent anions such as sulfat

At the start of the second decade of 21th century, the time has come to make the Smart Houses a reality for regular use. The different parts of a Smart House are researched but there are still distances from an applicable system, using the modern technology. In this paper we present an overview of the Smart House subsystems necessary for controlling the house using a mobile application efficiently and securely. The sequence diagram of the mobile application connecting to the server application and also the use-cases possible are presented. The challenges faced in designing the mobile application and illustrating the updated house top plane view in that application, are discussed and solutions are adapted for it. Finally the designed mobile application was implemented and the important sections of it were described, such as the interactive house top view map which indicates the status of the devices using predefined icons. The facilities to manage the scheduled tasks and defined rules are also implemented in this mobile application that was developed for use in Windows Mobile platform. This application has the capability of connecting to the main server using GPRS mobile internet an