## Learning Depth With Very Sparse Supervision

Existing works for depth estimation require massive amount of annotated training data or some form of hard-coded geometrical constraint . We train a specialized global-local networkarchitecture with what would be available to a robot interacting with theenvironment . From a pair of consecutive images, our proposed networkoutputs a latent representation of the observer’s motion between the images and a dense depth map .…

## A general framework for scientifically inspired explanations in AI

Explainability in AI is gaining attention in the computer science community in response to the increasing success of deep learning and the need to justify how such systems make predictions in life-critical applications . Authors claim current explanations in AI do not take into account the complexity of human interaction to allow foreffective information passing to not-expert users .…

## PhoBERT Pre trained language models for Vietnamese

PhoBERT is the first public large-scale monolingual language models pre-trained for Vietnamese . It outperforms XLM-R (Conneau et al., 2020) andimproves the state-of-the-art in multiple Vietnamese-specific NLP tasks . The models are available at https://://github.com/VinAIResearch/PhoberT .…

## PPMC RL Training Algorithm Rough Terrain Intelligent Robots through Reinforcement Learning

The Path Planning and Motion Control (PPMC) Training Algorithm is coupled with any generic reinforcement learning algorithm to teach robots how to respond to user commands and to travel to designated locations on a single neural network . The algorithm worksindependent of the robot structure, demonstrating that it works on a wheeledrover in addition the past results on a quadruped walking robot .…

## Batch Stationary Distribution Estimation

We consider the problem of approximating the stationary distribution of anergodic Markov chain given a set of sampled transitions . We develop a variational powermethod (VPM) that provides provably consistent estimates under general conditions . VPM yields significantly betterestimates across a range of problems, including queueing, stochasticdifferential equations, post-processing MCMC, and off-policy evaluation.…

## Knowledge Graphs on the Web an Overview

Knowledge Graphs are an emerging form of knowledge representation . In aknowledge graph, entities in the real world and/or a business domain are represented as nodes, which are connected byedges representing the relations between those entities . While companies such as Google, Microsoft, and Facebook have their own, non-public knowledge graphs, there is also a larger body of publicly available knowledge graphs such asDBpedia or Wikidata .…

## Using Image Captions and Multitask Learning for Recommending Query Reformulations

Interactive search sessions often contain multiple queries, where users submit a reformulated version of the previous query in response to the original results . We aim to enhance the query recommendation experience for a commercial image search engine. Our proposed methodology incorporates current state-of-the-art practices from relevant literature .…

## UFTR A Unified Framework for Ticket Routing

Corporations today face increasing demands for the timely and effectivedelivery of customer service . This creates the need for a robust and accurateautomated solution to what is formally known as the ticket routing problem . This task is to match each unresolved service incident, or “ticket”, to theright group of service experts .…

## Deep Image Spatial Transformation for Person Image Generation

Pose-guided person image generation is to transform a source person image to a target pose . This task requires spatial manipulations of source data . Convolutional Neural Networks are limited by the lack of ability to transform the inputs . In this paper, we propose a differentiableglobal-flow local-attention framework to reassemble the inputs at the featurelevel .…

## Out of Distribution Generalization via Risk Extrapolation REx

A weak form of out-of-distribution (OoD) generalization is the ability to successfully interpolate between multipleobserved distributions . In pursuit of strong OoD generalization, we introduce the principle of Risk Extrapolation (REx) REx can be viewed as encouraging robustness overaffine combinations of training risks, by encouraging strict equality betweentraining risks .…

## Detection and Mitigation of Bias in Ted Talk Ratings

This paper quantifies implicit bias in viewer ratings of TEDTalks, a social platform assessing social and professional performance . Unbiased data collection is essential to guaranteeing fairness in artificialintelligence models . In our paper, we presentstrategies to detect and mitigate bias that are critical to removing unfairness in AI .…

## Permutohedral GCN Graph Convolutional Networks with Global Attention

Graph convolutional networks update a node’s feature vector by aggregating features from its neighbors in the graph . This ignores potentially useful contributions from distant nodes . Identifying useful distant contributions is challenging due to scalability issues and oversmoothing . We introduce a globalattention mechanism where a node can selectively attend to, and aggregatefeatures from, any other node .…

## Sparsity Meets Robustness Channel Pruning for the Feynman Kac Formalism Principled Robust Deep Neural Nets

Deep neural nets (DNNs) compression is crucial for adaptation to mobiledevices . We use relaxed augmented Lagrangian based algorithms to prune the weightsof adversarially trained DNNs, at both structured and unstructured levels . The code is available at\url{https://://://www.gong.com/BaoWangMath/rvsm-rgsm-admm . We can atleast double the channel sparsity of the ResNet20 for CIFAR10 classification, meanwhile, improve the natural accuracy by $8.69$\% and the robust accuracy under the benchmark $20$ iterations of IFGSM attack by $5.42$\%.…

## Piecewise Linear Valued Constraint Satisfaction Problems with Fixed Number of Variables

Many combinatorial optimisation problems can be modelled as valued constraintsatisfaction problems . In this paper, we present a polynomial-time algorithmsolving the valued constraint satisfaction problem for a fixed number ofvariables . Our algorithm finds theinfimum of a piecewise linear function and decides whether it is a properminimum .…

## Hardness of Sparse Sets and Minimal Circuit Size Problem

We develop a polynomial method on finite fields to amplify the hardness ofspare sets in nondeterministic time complexity classes on a randomized streaming model . We also show that if MCSP is $ZPP$-hardunder polynnomineable time truth-table reductions, then $EXP\not=ZPP$. We also develop a method on a finite field that amplifies hardness of spare sets .…

## Descriptive complexity of real computation and probabilistic independence logic

We introduce a novel variant of BSS machines called Separate Branching BSSmachines . We develop a Fagin-type logical characterisation for languages decidable in non-deterministic polynomial time by S-BSS machines . We establish that on Boolean inputs NP onS-BSs machines without real constants characterises a natural fragment of thecomplexity class existsR (a class of problems polynnomine time reducible to the true existential theory of the reals) and hence lies between NP and PSPACE .…

## A Hybrid Lagrangian Eulerian Method for Topology Optimization

LETO is a new hybrid Lagrangian-Eulerian method for topologyoptimization . It uses a hybrid particle-grid MaterialPoint Method to solve for elastic force equilibrium . LETO’s objective achieves an average quantitative improvement of 20% in 3D and 2% in 2D . It unifies the treatment for both linear and non-linearelastic materials .…

## Modified Bee Colony optimization algorithm for computational parameter identification for pore scale transport in periodic porous media

Modified Bee Colonyalgorithm (MBC) based on a particular intelligent behavior of honeybee swarms . The algorithm was checked in a few benchmarks like Shekel, Rozenbroke,Himmelblau and Rastrigin functions . The proposed identification approach is applicable for different geometries (random and periodic) and for a range of process parameters .…

## PF Net Point Fractal Network for 3D Point Cloud Completion

Point Fractal Network (PF-Net) is a novel learning-based approach for precise and high-fidelity point cloud completion . Unlike existing networks, PF-Net preserves thespatial arrangements of the incomplete point cloud and can figure out the geometrical structure of the missing region(s) in the prediction .…

## MINA Convex Mixed Integer Programming for Non Rigid Shape Alignment

We present a convex mixed-integer programming formulation for non-rigid shapematching . We propose a novel shape deformation model based on anefficient low-dimensional discrete model . Our approach combines severalfavourable properties: it is independent of the initialisation, it is much more efficient to solve to global optimality compared to analogous quadraticassignment problem formulations .…

## BitcoinF Achieving Fairness for Bitcoin in Transaction Fee Only Model

A fair blockchain is expected to have healthy metrics; highhonest mining power, low processing latency, i.e., low wait times for transactions and stable price of consumption . As Bitcoin matures, the influx of transactions increases and the block rewards become insignificant .…

## A Permutation Equivariant Neural Network Architecture For Auction Design

Designing an incentive compatible auction that maximizes expected revenue is a central problem in Auction Design . Building on the success of deeplearning, a new approach was recently proposed by Duetting et al. inwhich the auction is modeled by a feed-forward neural network and the designproblem is framed as a learning problem .…

## Constant delay enumeration with FPT preprocessing for conjunctive queries of bounded submodular width

Marx (STOC~2010, J.~ACM 2013) introduced the notion of submodular width of aconjunctive query (CQ) For non-Boolean queries, the size of the query result may be far too large to becomputed entirely within FPT time . We have to tackle the additional technicaldifficulty to ensure that these can be enumerated efficiently .…

## Scaling Up Multiagent Reinforcement Learning for Robotic Systems Learn an Adaptive Sparse Communication Graph

Complexity of multiagent reinforcement learning (MARL) increases exponentially with respect to the agent number . Thisscalability issue prevents MARL from being applied in large-scale multiagentsystems . We propose anadaptive sparse attention mechanism by generalizing a sparsity-inducingactivation function . Then a sparse communication graph in MARL is learned bygraph neural networks based on this new attention mechanism .…

## AutoPhase Juggling HLS Phase Orderings in Random Forests with Deep Reinforcement Learning

The performance of the code a compiler generates depends on the order in which it applies the optimization passes . Choosing a good order–often referred to as the phase-ordering problem, is an NP-hard problem . In our evaluation, we show that AutoPhase improves circuit performance by 28% when compared to using the -O3 compiler flag, and achieves competitive results compared to the state-of-the-art solutions, while requiring fewersamples .…

## Learning to Simulate Human Movement

Modeling how human moves on the space is useful for policy-making intransportation, public safety, and public health . The human movements can beviewed as a dynamic process that human transits between states (e.g.,locations) over time . In experiments on real-world datasets, wedemonstrate that the proposed method can achieve superior performance against the state-of-the-art methods in predicting the next state and generating long-term future states .…

## Stein Variational Inference for Discrete Distributions

The proposed method outperforms traditional algorithms such as Gibbs sampling and discontinuous Hamiltonian Monte Carlo on various benchmarks of discrete graphical models . We demonstrate that our method provides a promising tool for learning ensembles of binarized neuralnetwork (BNN), outperforming other widely used ensemble methods on learningbinarized AlexNet on CIFAR-10 dataset .…

## Securing of Unmanned Aerial Systems UAS against security threats using human immune system

An Intrusion Detection System (IDS) has been proposed in the proposed design to protect against security problems using the human immune system (HIS) The IDSs are used to detect andrespond to attempts to compromise the target system . In themapping, insecure signals are equivalent to an antigen that are detected byantibody-based training patterns and removed from the operation cycle .…

## Differentially Private Deep Learning with Smooth Sensitivity

Ensuring the privacy of sensitive data used to train modern machine learning models is of paramount importance in many areas of practice . We propose a novel voting mechanism with smooth sensitivity, which we call Immutable Noisy ArgMax, that can bear very large random noising from the teacher without affecting the useful information transferred to the student .…

## MonoPair Monocular 3D Object Detection Using Pairwise Spatial Relationships

Monocular 3D object detection is an essential component in autonomous driving while challenging to solve, especially for those occluded samples . The proposed detector computes uncertainty-aware predictions for object locations and 3D distances for the adjacent object pairs, which are subsequently jointlyoptimized by nonlinear least squares .…

## 1D CNN Based Network Intrusion Detection with Normalization on Imbalanced Data

Intrusion detection system (IDS) plays an essential role in computer networksprotecting computing resources and data from outside attacks . Recent IDS faces challenges improving flexibility and efficiency of the IDS for unexpected andunpredictable attacks . We propose a deep learning approach for developing the efficientand flexible IDS using one-dimensional Convolutional Neural Network (1D-CNN) The method can be used for supervised learning on time-series data .…

## GPM A Generic Probabilistic Model to Recover Annotator s Behavior and Ground Truth Labeling

In the big data era, data labeling can be obtained through crowdsourcing . But the obtained labels are generally noisy, unreliable or even adversariesarial . In this paper, we propose a probabilistic graphical annotationmodel to infer the underlying ground truth and annotator’s behavior .…

Existing methods tend to overfit training data in seen environments and fail to generalize well in previously unseen environments . We aim at learning ageneralized navigation model from two novel perspectives . We introduce amultitask navigation model that can be seamlessly trained on bothVision-Language Navigation (VLN) and Navigation from Dialog History (NDH) tasks .…

## Differential Evolution with Individuals Redistribution for Real Parameter Single Objective Optimization

Differential Evolution (DE) is quite powerful for real parameter single objective optimization . We propose a new flow of DE, termed DE with individualsredistribution, in which a process of individuals redistribution will be calledwhen progress on fitness is low for generations .…

## A Study on Multimodal and Interactive Explanations for Visual Question Answering

Explainable AI (XAI) approaches aimat mitigating the lack of transparency in deep networks . Evidence of the effectiveness of these approaches in improving usability, trust, and understanding of AI systems are still missing . We evaluate multimodalexplanations in the setting of a Visual Question Answering (VQA) task, by asking users to predict the response accuracy of a VQA agent with and withoutexplanations .…

## Towards Automatic Face to Face Translation

In light of recent breakthroughs in automatic machine translations, we propose a novel approach that we term as “Face-to-FaceTranslation” LipGAN is a novel visual module for generating realistic talking faces from the translated audio . We argue that there is a need for systems that can automatically translate a video of a person speaking in language A into a target language B with realistic lipsynchronization .…

## Data Pre Processing and Evaluating the Performance of Several Data Mining Methods for Predicting Irrigation Water Requirement

Recent drought and population growth are planting unprecedented demand for water resources . Irrigated agriculture is one of the major consumers of freshwater . A large amount of water in irrigated areas is wasted due to poor water management practices .…

## Solving Satisfiability of Polynomial Formulas By Sample Cell Projection

A new algorithm for deciding satisfiability of polynomial formulas over the reals is proposed . The key point of the algorithm is a new projection operator called sample-cell projection operator . It is custom-made forConflict-Driven Clause Learning (CDCL)-style search . Experiments show the effectiveness of the new algorithm .…

## Learn Task First or Learn Human Partner First Deep Reinforcement Learning of Human Robot Cooperation in Asymmetric Hierarchical Dynamic Task

The deep reinforcement learning method for human-robot cooperation (HRC) is promising for its high performance when robots are learning complex tasks . The applicability of such an approach in a real-world context is limited due to long training time, additional training difficulty caused byinconsistent human performance .…

## Deep Active Learning for Biased Datasets via Fisher Kernel Self Supervision

Active learning aims to minimize labeling efforts for data-demanding deep neural networks by selecting the most representative data points . Currently used methods are ill-equipped to deal with biased data . We propose a low-complexity method forfeature density matching using self-supervised Fisher kernel (FK) as well as novel pseudo-label estimators .…

## Fine grained Video Text Retrieval with Hierarchical Graph Reasoning

The Hierarchical Graph Reasoning (HGR) model decomposes video-text matching into global-to-local levels . Attention-based graph reasoning is utilized to generate hierarchicaltextual embeddings . The HGR model aggregates matchings from differentvideo-text levels to capture both global and local details . The model also enables better generalization acrossdatasets and improves the ability to distinguish fine-grained semanticdifferences, such as fine-granized semantic differences, according to the authors of this article .…

## Say As You Wish Fine grained Control of Image Caption Generation with Abstract Scene Graphs

Most image captioning models are intention-agnostic which generate diverse descriptions according to different user intentions . In this work, we propose the Abstract Scene Graph (ASG) structure to represent user intention in fine-grained level . We propose a novel ASG2Caption model, which is able to recognise user intentions and .…

## Participatory Budgeting Models and Approaches

Participatory budgeting is a democratic approach to deciding the funding of public projects . It has been adopted in many cities across the world . We present a mathematical model for participatorybudgeting . We survey various approaches andmethods from the literature, giving special emphasis on issues of preferenceelicitation, welfare objectives, fairness axioms, and voter incentives .…

## Simple Mechanisms for Non linear Agents

We quantify extent to which posted pricing approximatelyoptimizes welfare and revenue for a single agent . We give a reduction framework that extends the approximation of multi-agent pricing-based mechanisms from linear utility to nonlinear utility . This reduction framework is broadlyapplicable as Alaei et al.…

## Profinite congruences and unary algebras

Profinite congruences on profinite algebras determining profinite quotientsare difficult to describe . We establish an adjunction between profinite unary algebra and profinite monoids . We also show that the Polish representation of the free profinite algebra is faithful . We show that our conjecture fails for unary algebnas and that closed congruence on relatively free profininite semigroups are not necessarilyprofinite .…

## Process algebra process scheduling and mutual exclusion

Multi-threading is interleaved according to what is known as a process-scheduling policy in the field of operating systems . In the current paper, wedo so with the variant of ACP known as ACP$_\epsilon . The choice ofACP$_$___$_.\eppsilon\$ stems from the need to cover more process-schedule policies .…

## Asynchronous Policy Evaluation in Distributed Reinforcement Learning over Networks

This paper proposes a scheme for policy evaluation of distributed reinforcement learning (DisRL) over peer-to-peer networks . Numerical experiments show that our method speeds up linearly w.r.t. the number of nodes, and is robust tostraggler nodes . To the best of our knowledge, our work is the firsttheoretical analysis for .…

## Weak Texture Information Map Guided Image Super resolution with Deep Residual Networks

Single image super-resolution (SISR) is an image processing task which obtained high-resolution images from a low-resolution image . Deep learning methodshave brought important crucial improvement for SISR . However, we observe that no matter how deeper the networks are designed, they usually do not have good generalization ability .…

## Demonstrating Immersive Media Delivery on 5G Broadcast and Multicast Testing Networks

This work presents eight demonstrators and one showcase developed within the 5G-Xcast project . They experimentally demonstrate and validate key technicalenablers for the future of media delivery, associated with multicast andbroadcast communication capabilities in 5th Generation (5G) The demonstrations cover use cases belonging to two verticals: Media & Entertainmentand Public Warning .…

## Harmonics Based Representation in Clarinet Tone Quality Evaluation

Music tone quality evaluation is generally performed by experts . We present a new method for identifying the clarinet reed quality by evaluating tone quality based on the harmonic structure and energy distribution of the harmonics . The resultsshow that our new method .…