Transfer Learning for Human & AI

Datasets for training crowd counting deep networks are typically heavy-tailedin count distribution . De facto statistical measures exhibit large variance and tend to be unreliable indicators of performance across the count range . To enable principled andbalanced minibatch sampling, we propose a novel smoothed Bayesian samplestratification approach .…

In recent years, many recommender systems using network embedding (NE) suchas graph neural networks (GNNs) have been extensively studied in the sense ofimproving recommendation accuracy . SiReN has three key components: constructing a signed bipartite graph for moreprecisely representing users’ preferences, splitting into twoedge-disjoint graphs with positive and negative edges each, generating twoembeddings for the partitioned graphs .…

We use a greedy strategy to list the spanning trees of the fan graph, $F_n$, such that successive trees differ by pivoting a single edge around a vertice . The resulting listing is then studied to find a recursive algorithm that produces the same listing in $O(1)$-amortized time .…

MARS (Multi-channel Ambiently-poweredRealtime Sensing) enables battery-free sensing and wireless communication . Multiple MARS systems can operate in the same environment bytuning each oscillator circuit to a different frequency range . The nanowattpower budget allows the system to be powered directly through ambient energysources like photodiodes or thermoelectric generators .…

The relative interlevel set cohomology (RISC) is an invariant of real-valuedcontinuous functions . It is closely related to the Mayer–Vietoris pyramid introduced by Carlsson, de Silva, and Morozov . We provide a structure theorem, which applies to the RISC if it is pointwise finite dimensional (pfd) or,equivalently, $q$-tame .…

Generalized Geography is a combinatorial game played on a directed graph . Players take turns moving a token from vertices to vertices . A player unable to move loses . The computational complexity of determining which player should win from a given position is PSPACE-complete .…

Convolutional neural networks (CNNs) have become the state-of-the-art tool for dealing with unsolved problems in computer vision and image processing . PowerLinear activation functions are based on the polynomial kernel generalization of the convolution operator . They are used in CNN models to classify the inversion of grayscale images as accurately as the original graycale images, which is significantly better than commonly used activation functions .…

We study edit distance computation with preprocessing: the preprocessingalgorithm acts on each string separately, and then the query algorithm takes as input the two preprocessed strings . This model is inspired by scenarios wherewe would like to compute edit distance between many pairs in the same pool ofstrings .…

Ternary logic system is the most promising and pursued alternate to theprevailing binary logic systems due to the energy efficiency of circuits . In this paper, we have proposed a ternary 3-Transistor Dynamic Random-Access Memory (3T-DRAM) cell using a single word-line for both read and write operation .…

The literal and initial literal shuffle have been introduced to model the behavior of two synchronized processes . However, it is not possible to describe the synchronization of multiple processes . Here, we extend the literal shuffle and theinitial literal shuffle to multiple arguments .…

In the absence of a central controller, agents communicate information via atime-varying communication network while preserving privacy . We prove the convergence of all the 3 MAN algorithms to a globally asymptotically stablepoint of the ODE corresponding to the actor update .…

The expected metaverseshould be a realistic society with more direct and physical interactions, while the concepts of race, gender, and even physical disability would be weakened . However, the development of metaverse is still in its infancy, with great potential for improvement .…

Most deep-learning frameworks for understanding biological swarms are designed to fit perceptive models of group behavior to individual-level data . Despite considerable advances in automated tracking, these methods are still very expensive or unreliable when tracking large numbers of animals simultaneously .…

Matrix factorizations are among the most important building blocks of computing . State-of-the-art libraries, however, are notcommunication-optimal, underutilizing current parallel architectures . Wepresent novel algorithms for Cholesky and LU factorizations that utilize anasymptotically communication-optimistic 2.5D decomposition . Our code is ScaLAPACK-compatible and available as an open-source library.…

Most deep-learning frameworks for understanding biological swarms are designed to fit perceptive models of group behavior to individual-level data . Despite considerable advances in automated tracking, these methods are still very expensive or unreliable when tracking large numbers of animals simultaneously .…

We study edit distance computation with preprocessing: the preprocessingalgorithm acts on each string separately, and then the query algorithm takes as input the two preprocessed strings . This model is inspired by scenarios wherewe would like to compute edit distance between many pairs in the same pool ofstrings .…

This paper presents an effective method for generating a spatiotemporal(time-varying) texture map for a dynamic object using a single camera . The input of our framework is a 3D template model and an RGB-D image sequence . Weformulate the problem as an MRF optimization and define cost functions to construct a plausible texture for a .…

Flow aggregation, namely origins and destinationsaggregated simultaneously, is one of the most common patterns in cities . The movement of humans and goods in cities can be represented by constrainedflow, defined as the movement of objects between origin anddestination in road networks .…

Generalized Geography is a combinatorial game played on a directed graph . Players take turns moving a token from vertices to vertices . A player unable to move loses . The computational complexity of determining which player should win from a given position is PSPACE-complete .…

An adaptive interpolation scheme is proposed to accurately calculate thewideband responses in electromagnetic simulations . In the proposed scheme, thesampling points are divided into several groups based on their groups . If the relative error does not satisfy the predefined threshold in a specific frequency band, it will be refined until the error criteria is met .…

An adaptive interpolation scheme is proposed to accurately calculate thewideband responses in electromagnetic simulations . In the proposed scheme, thesampling points are divided into several groups based on their groups . If the relative error does not satisfy the predefined threshold in a specific frequency band, it will be refined until the error criteria is met .…

We design a new efficient strategy synthesis method applicable to adversarialpatrolling problems on graphs with arbitrary-length edges . The core ingredient is an efficient algorithm for computing the value and the gradient of a function assigning to every strategy its “protection” achieved .…

Flow aggregation, namely origins and destinationsaggregated simultaneously, is one of the most common patterns in cities . The movement of humans and goods in cities can be represented by constrainedflow, defined as the movement of objects between origin anddestination in road networks .…

We study edit distance computation with preprocessing: the preprocessingalgorithm acts on each string separately, and then the query algorithm takes as input the two preprocessed strings . This model is inspired by scenarios wherewe would like to compute edit distance between many pairs in the same pool ofstrings .…

In this paper, we study the privacy-preserving task assignment in spatialcrowdsourcing, where the locations of both workers and tasks, prior to their release to the server, are perturbed with Geo-Indistinguishability . We propose the k-Switch solution, which first divides the workers into small groups based on the perturbed distance between workers/tasks, and then utilizes Homomorphic Encryption (HE) based secure computation to enhancethe task assignment .…

Generalized Geography is a combinatorial game played on a directed graph . Players take turns moving a token from vertices to vertices . A player unable to move loses . The computational complexity of determining which player should win from a given position is PSPACE-complete .…

Matrix factorizations are among the most important building blocks of computing . State-of-the-art libraries, however, are notcommunication-optimal, underutilizing current parallel architectures . Wepresent novel algorithms for Cholesky and LU factorizations that utilize anasymptotically communication-optimistic 2.5D decomposition . Our code is ScaLAPACK-compatible and available as an open-source library.…

In this work, we aim to calibrate the score outputs of an estimator for the binary classification problem by finding an ‘optimal’ mapping to classprobabilities . We show that for the given target variables and the score outcomes of an .…

In this paper, we propose a learned video codec with a residual predictionnetwork (RP-Net) and a feature-aided loop filter . For the RP-Net, we exploit the residual of previous multiple frames to further eliminate the redundancy of the current frame residual .…

We propose a class of functional dependencies for temporal graphs calledTGFDs . TGFDs capture both attribute-value and topologicalstructures of entities over a valid period of time in a temporal graph . We show that thesatisfiability and validation problems are coNP-complete and the implicationproblem is NP-complete .…

Virtual Reality (VR) games that feature physical activities have been shownto increase players’ motivation to do physical exercise . For suchexercises to have a positive healthcare effect, they have to be repeated several times a week . We propose to use experience-driven Procedural ContentGeneration for DDA in VR exercise games by procedurally generating levels thatmatch the player’s current capabilities .…

Musical instruments recognition is a widely used application for musicinformation retrieval . Most of previous musical instruments recognitiondataset focus on western musical instruments . This paper propose a traditional Chinese music dataset for training model and performance evaluation, named ChMusic .…

We consider the problem of automatically proving resource bounds . We show that proving bounds in any such decomposition yields asound resource bound in the original program . We present a framework for selectively-amortized analysis that mixes worst-caseand amortized reasoning via a property decomposition and a programtransformation .…

Monitoring player impacts is vitally important tounderstanding and protecting from injuries like concussion . Traditional machine learning approaches fail to detect true positives and prevent falsenegatives . In this work we present a new physics-informed machine learning model that can be used to analyze kinematic data from an instrumented mouthguard anddetect impacts to the head .…

Repairs are consistent instances that minimally differ from the original inconsistent instance . Consistent answers to a query from a possibly inconsistent database are answers that are simultaneously retrieved from every possible repair of the database . In this paper we show how to use the repair programs to transform the problem ofconsistent query answering into a problem of reasoning w.r.t.…

The Episode Matching Problem is to compute thelength of the shortest substring of $S$ that contains $P$ as a subsequence . The best known upper bound for this problem is $O(nm)$ by Das et al. (1997) The problem is well studied and has many applications in data mining .…

A new proof method, already used in arXiv:2104.14316, P not equal NP can be proved . A time limit is set for an arbitrary Turing machine and an input word is rejected on a timeout . Due to thehalting problem, whether a word is accepted can only be determined at runtime .…

We present an approach for integrating the time evolution of quantum systems . Our PARAllelized Matrix Exponentiation for Numerical Timeevolution (PARAMENT) implementation runs on CUDA-enabled graphics processing units . The algorithm can largely be implemented using the recently-specified batchedversions of the BLAS routines, and can therefore be easily ported to a variety of platforms .…

Traditional resilient systems operate on fully-replicated fault-tolerant clusters, which limits their scalability and performance . One way to make the step towards resilient high-performance systems that can deal with hugeworkloads, is by enabling independent fault tolerantant clusters to efficientlycommunicate and cooperate with each other .…

In this work, we aim to calibrate the score outputs of an estimator for the binary classification problem by finding an ‘optimal’ mapping to classprobabilities . We show that for the given target variables and the score outcomes of an .…

The Episode Matching Problem is to compute thelength of the shortest substring of $S$ that contains $P$ as a subsequence . The best known upper bound for this problem is $O(nm)$ by Das et al. (1997) The problem is well studied and has many applications in data mining .…

The performance of a computer vision model depends on the size and quality of its training data . We propose an AU-Calibrated FacialExpression Recognition (AUC-FER) framework that utilizes facial action units(AUs) and incorporates the triplet loss into the objective function .…

A new proof method, already used in arXiv:2104.14316, P not equal NP can be proved . A time limit is set for an arbitrary Turing machine and an input word is rejected on a timeout . Due to thehalting problem, whether a word is accepted can only be determined at runtime .…

Monitoring player impacts is vitally important tounderstanding and protecting from injuries like concussion . Traditional machine learning approaches fail to detect true positives and prevent falsenegatives . In this work we present a new physics-informed machine learning model that can be used to analyze kinematic data from an instrumented mouthguard anddetect impacts to the head .…

Thom Fr\”uhwirth presented a short, elegant and efficient Prolog program for the n queens problem . However the program may be seen as rather tricky and onemay not be convinced about its correctness . This paper explains the program in a declarative way, and provides proofs of its correctness and completeness .…

We study the canonical weak distributive law of the powerset monadover the semimodule monad for a certain class of semirings containing, inparticular, positive semifields . For this subclass we characterise $delta$ as a convex closure in a set . Using the abstract theory ofweak distributive laws, we compose the monad of convex subsets of the free semimmodule .…

Lamport’s celebrated Paxos consensus protocol is generally viewed as a hard-to-understand algorithm . We take a step towards automatically proving the safety of Paxos by taking advantage of three structural features in its specification . We note that these structural features are not specific to Paxos and that IC3PO can serve as an automatic general-purpose protocol verification tool .…

Program synthesis has seen many new applications in recent years, in largepart thanks to the introduction of SyGuS . We introduce a multi-phasealgorithm for the synthesis of recursion programs . We solve constraints individually and treat them as examples of how a recursion program would behave .…

Federated Learning (FL) is a privacy-protected machine learning paradigm that allows model to be trained directly at the edge without uploading data . One of the biggest challenges faced by FL in practical applications is theheterogeneity of edge node data .…

Efficient frequency-domain Full Waveform Inversion (FWI) of long-offset/wide-azimuth node data can be designed with a few discreterequencies . However, 3D frequency- domain seismic modeling remains challengingsince it requires solving a large and sparse linear indefinite system perfrequency . When such systems are solved with direct methods or hybriddirect/iterative solvers, based upon domain decomposition preconditioner,finite-difference stencils on regular Cartesian grids should be designed to conciliate compactness and accuracy, the former being necessary to mitigate thefill-in induced by the Lower-Upper (LU) factorization .…