Transfer Learning for Human & AI

Proactive tile-based virtual reality (VR) video streaming can use the traceof FoV and eye movement to predict future requested tiles, then renders and delivers the predicted tiles before playback . Quality of experience (QoE) depends on the combined effect of tile prediction and consumed resources .…

We study combinatorial inequalities for various classes of set systems . We use the language formulation of greedoids which allows a linear algebraic setup . The underlying non-commutative nature of matrices associated withgreedoids allows us to proceed beyond polymatroids and prove the equalityconditions .…

Privacy is of worldwide concern regarding activities and processes thatinclude sensitive data . Artificial intelligenceareas, such as machine learning and natural language processing, have alreadysuccessfully employed privacy-preserving mechanisms in order to safeguard dataprivacy in a vast number of applications . Information retrieval (IR) is prone to privacy threats such as attacks and unintended disclosures of documents and search history, which may cripple the security of users and bepenalized by data protection laws .…

Internet of things (IoT) has significantly altered the traditional lifestyle to a highly technologically advanced society . Some of the significanttransformations that have been achieved through IoT are smart homes, smarttransportation, smart city, and control of pollution . A considerable number of studies have been conducted and continue to be done to increase the use of technology through IoT .…

Python has become a dominant programming language for emerging areas likeMachine Learning (ML), Deep Learning (DL), and Data Science (DS) An attractive feature of Python is that it provides easy-to-use programming interface while allowing library developers to enhance performance of their applications .…

An emerging problem in themodern era is fake news detection — many easily available pieces of information are not necessarily factually correct . In this work we explore how differentdocument representations, ranging from simple symbolic bag-of-words, to contextual, neural language model-based ones can be used for efficient fakenews identification .…

In this work, we introduce the pattern-domain pilot design paradigm based on a “superposition of orthogonal-building-blocks” with significantly largercontention space to enhance the massive machine-type communications (mMTC)random access (RA) performance in massive multiple-input multiple-output (MIMO) systems . The increased contention space, i.e.,…

In reversible computations one is interested in the development of mechanisms allowing to undo the effects of executed actions . In this paper, we consider theproblem of reversing the effect of the execution of groups of actions (steps) Using Petri nets as a system model, we introduce concepts related to this newscenario .…

With the advancement of computer technology, there is a surge of interest ineffective mapping methods for objects in higher-dimensional spaces . In this work, we develop a unifying framework for computing $n$-dimensional quasi-conformal mappings . We demonstrate the effectiveness of the proposed framework using various experiments in two- and three-dimensions, with applications to medical image registration, adaptiveremeshing and shape modeling .…

In this paper, we present a general framework that allows us to analyze various allocation processes that slightly prefer allocating intounderloaded, as opposed to overloaded bins . We consider the sequential allocation of $m$ balls (jobs) into $n$ bins(servers) by allowing each ball to choose from some bins sampled uniformly atrandom .…

Normal estimation on 3D point clouds is a fundamental problem in 3D vision and graphics . Current methods often show limited accuracy in predicting normals at sharp features (e.g., edges and corners) and less robustness to noise . We are motivated that local patches on isotropic and anisotropic surfaces have similar or distinct normals .…

In this paper, we present a general framework that allows us to analyze various allocation processes that slightly prefer allocating intounderloaded, as opposed to overloaded bins . We consider the sequential allocation of $m$ balls (jobs) into $n$ bins(servers) by allowing each ball to choose from some bins sampled uniformly atrandom .…

A long-standing criticism against neural network models is the lack of interpretability . Increasing interest in interpreting neural NLP models has spurred a diverse array of interpretation methods over recent years . We provide a comprehensive review of variousinterpretation methods for neural models in NLP .…

The task of talking head generation is to synthesize a lip synchronizedtalking head video by inputting an arbitrary face image and audio clips . Most existing methods ignore the local driving information of the mouth muscles . In this paper, we propose a novel recurrent generative network that uses bothaudio and speech-related facial action units as the driving information .…

The study demonstrates the feasibility of creating a federated NLP algorithmdevelopment and benchmarking platform to enhance multi-institution clinical NLP study . The generated corpora derived out of thetexts from multiple intuitions and gold standard annotated are tested on asingle institution’s rule set .…

Change captioning is to use a natural language sentence to describe the fine-grained disagreement between two similar images . Viewpoint change is a typical distractor in this task, because it changes the scale and location of the objects and overwhelms the representation of real change .…

The ever-increasing number of nodes in current and future wirelesscommunication networks brings unprecedented challenges for the allocation of available communication resources . This is caused by the combinatorial nature of the resource allocation problems, which limits the performance of state-of-the-art techniques when the network size increases .…

The formalism of quantum theory over discrete systems is extended in twosignificant ways . Tensors and traceouts are generalized, so that systemscan be partitioned according to almost arbitrary logical predicates . Second,quantum evolutions are generalized to act over network configurations, in sucha way that nodes be allowed to merge, split and reconnect coherently in asuperposition .…

In order to achieve deep natural language understanding, syntacticconstituent parsing is a vital step . We extend the framework of sequence-to-sequence models for constituent parsing . We design novel linearizationsthat can fully produce discontinuities and, for the first time, we test asequence to-sequence model on the main discontinuous benchmarks, obtainingcompetitive results on par with task-specific discontinuous constituent parsers .…

The QuantumApproximate Optimization Algorithm (QAOA) has been the subject of many studies . We partially address these issues for a specific combinatorial optimization problem: diluted spin models . We validate our results with numerical experiments suggesting they may have a larger reach than rigorously established .…

We describe the 2021 Key Point Analysis (KPA-2021) shared task on key pointanalysis that we organized as a part of the 8th Workshop on Argument Mining(ArgMining 2021) at EMNLP 2021 . We outline various approaches and discuss the results of the shared task .…

The multi-headmulti-modal DIN improves the recommendation prediction effect, and outperforms state-of-the-art methods on various comprehensive indicators . Researchers have optimized more on therecommendation model network structure, they have less research on enrichingrecommendation models, and there is still room for in-depth model optimization .…

In this paper, we study a matching market model on a bipartite network where agents on each side arrive and depart stochastically by a Poisson process . We design a mechanism that decides not only which agentsto match, but also when to match them, to minimize the expected number ofunmatched agents .…

Can every connected graph burn in $\lceil \sqrt{n} \rceil $ steps? While this remains open, we prove that it is asymptotically true when the graphis much larger than its growth . We prove that theconjecture for graphs of bounded growth boils down to a finite number of cases .…

Multi-agent reinforcement learning (MARL) algorithms have made promisingprogress in recent years . But existing MARL algorithms still suffer from the sample inefficiency problem . State-based episodic memory (SEM) is the first work to introduce EM into MARL . It has lower space complexity and time complexity than state and action based EM (SAEM), which is originally proposed for single-agent learning .…

Mobile edge computing (MEC) is a prominent computing paradigm which expandsthe application fields of wireless communication . The dynamics and complexities of content popularities over space and time as well as the privacy preservation of userspose significant challenges to EC optimization .…

In this paper, we propose to build a stylish image captioning model through aMulti-style Multi modality mechanism (2M) We demonstrate that with 2M, we can build an effective stylish captioner . We show how this 2M mechanism can be used to buildstylish captioning models and show how these models can be utilized to provide explanations of likely errors in the models .…

An aspect-guided graph pooling operator is proposed to extract the aspect-relevant information from thereview-based syntax graphs to model the user’s preferences on an item at theaspect level . Then, a hierarchical explanation decoder is developed to generate aspect-related explanations based on the attention mechanism .…

Self-attention models pre-trained on large amounts of unannotated data have been hugely successful when fine-tuned on downstream tasks from a variety of domains and languages . We build a singleencoder with the BERT objective on unlabeled text together with the w2v-BERT objective .…

Many-Fermion Dynamics-nuclear, or MFDn, is a configuration interaction (CI)code for nuclear structure calculations . It is a platform-independent Fortran90 code using a hybrid MPI+X programming model . For CPU platforms the application has a robust and optimized OpenMP implementation for shared memoryparallelism.…

Communication among devices in multi-GPU systems plays an important role interms of performance and scalability . In order to optimize an application,programmers need to know the type and amount of the communication happening among GPUs . In this work, we extend ComScribe to identify communication among GPUs for collective and P2P communicationprimitives in NVIDIA’s NCCL library .…

SocialVisTUM is a proposed interactive visualization toolkit . It displays a graph with topics asnodes and their correlations as edges . Further details are displayedinteractively to support the exploration of large text collections . The toolkit optimizes automatically on custom data for optimalcoherence.…

We evaluate a simple approach to improving zero-shot multilingual transfer ofmBERT on social media corpus by adding a pretraining task called translationpair prediction . Our approach assumes access to translations (exact or approximate) between source-target language pairs, where we fine-tune a modelon source language task data and evaluate the model in the target language .…

This paper proposes a technique for adding a new source or target language to an existing multilingual NMT model without re-training it on the initial set of languages . It consists in replacing the shared vocabulary with a smalllanguage-specific vocabulary and fine-tuning the new embeddings on the new language’s parallel data .…

We consider the problem of multilingual unsupervised machine translation,translating to and from languages that only have monolingual data by using auxiliary parallel language pairs . We propose instead to use denoising adapters, adapter layers, on top of pre-trained mBART-50 . The resulting translations are on-par with back-translated as measured by BLEU, and further it allows adding unseen languages incrementally .…

Large-scale pretrained language models have been shown to learneffective linguistic representations for many NLP tasks . We propose a simple approach to incorporate speaker social context into the learnedrepresentations of large-scale language models . We evaluate our approach on geographically-sensitivelanguage-modeling tasks and show a substantial improvement (more than 100% relative lift on MRR) compared to baselines .…

Recent progress in pretrained Transformer-based language models has shown success in learning contextual representation of text . However, due to the quadratic self-attention complexity, most of the pretrained Transformers models can only handle relatively short text . In this work, we propose to use a graphattention network on top of the available .…

Weighted voting is a conventional approach to improving the performance of systems based on commonly-used majority quorum systems in heterogeneous environments . In long-lived systems, a weight reassignment protocol is required to reassign weights over time in order to accommodate performance variations accordingly .…

We develop methodology for three-dimensional (3D) radial visualization of multidimensional datasets . We show that this uniform distribution provides the best visualization with minimal artificial visualcorrelation for data with uncorrelated variables . We use RadViz3D to illustrate (i) the chemical composition of Longquan celadon ceramics and their Jingdezhen imitation overcenturies, and (ii) US regional SARS-Cov-2 variants’ prevalence in the Covid-19pandemic during the summer 2021 surge of the Delta variant .…

This paper investigates multiuser multi- input single-input single-output downlinksymbiotic radio communication systems . The proposed scheme significantly improves the average sum-rate of the primary system, while guaranteeing the decodingperformance of the secondary system . We exploit the Schur complement thatfacilitates the design of a suboptimal beamforming algorithm based on convex approximation .…

We present a novel characterization of the mapping of multiple parallelismforms onto hierarchical accelerator system . We experimentally verify the substantial effect of these mappings onall-reduce performance (up to 448x) We offer a novel syntax-guided programsynthesis framework that is able to decompose reductions over one or moreparallelism axes to sequences of collectives in a hierarchy- and mapping-awareway .…

Asynchronous iterations are more and more investigated for both scaling and fault-resilience purpose on high performance computing platforms . This paper advocates a novel application direction targeting time-decomposed time-parallel approaches . It turned out that Parareal andasync-Parareal feature very close convergence conditions, asymptoticallyequivalent, including the finite-time termination property .…

International discussions on the topic of how to define and define what a “safe enough” Automated Driving System is are currentlyhinged on the question of determining the relationship between “positive riskbalance” (PRB) and “absence of unreasonable risk” (AUR) The argumentation in this paper is aimed at showing that the two interpretations for PRB can actually complement each other, but can be considered independently, and can both besubsumed within non-prescriptive guidelines toward ADS safety assurance .…

Distributed energy resources (DERs) are growing rapidly and are reshaping power systems . Feed-in-tariff (FIT) is usually adopted by utilities to pay DER owners certain rates for supplying energy to the grid . An alternative to FIT is amarket-based approach; that is, consumers trade energy in anauction-based peer-to-peer (P2P) market .…

The study demonstrates the feasibility of creating a federated NLP algorithmdevelopment and benchmarking platform to enhance multi-institution clinical NLP study . The generated corpora derived out of thetexts from multiple intuitions and gold standard annotated are tested on asingle institution’s rule set .…

Applications making excessive use of single-object based data structures can see a drop in efficiency over a period of time due to the randomization of nodes in memory . This slow down isdue to the ineffective use of the CPU’s L1/L2 cache .…

In this paper, we propose sentiment classification models based on BERTintegrated with DRO (Distributionally Robust Classifiers) to improve model performance on datasets with distributional shifts . We added 2-Layer Bi-LSTM, projection layer (onto simplex or Lp ball), and linear layer on top of BERT toachieve distributionally robustness .…

Symbolic Interactive Language Grounding benchmark (SILG) unifies a collection of diverse grounded language learning environments . SILG consists of grid-world environments that require generalization to new dynamics, entities, and partially observed worlds (RTFM,Messenger, NetHack) and symbolic counterparts of visual worlds that require interpreting rich natural language with respect to complex scenes (ALFWorld, Touchdown) Together, these environments provide diverse groundingchallenges in richness of observation space, action space, languagespecification, and plan complexity .…

An uncertainty compiler is a tool that automatically translates original computer source code lacking explicit uncertainty analysis into code containing uncertainty representations and uncertainty propagation algorithms . It handles the specifications of input uncertainties and inserts callsto intrusive uncertainty quantification algorithms in the library .…

With the advancement of computer technology, there is a surge of interest ineffective mapping methods for objects in higher-dimensional spaces . In this work, we develop a unifying framework for computing $n$-dimensional quasi-conformal mappings . We demonstrate the effectiveness of the proposed framework using various experiments in two- and three-dimensions, with applications to medical image registration, adaptiveremeshing and shape modeling .…