Towards Trustworthy Predictions from Deep Neural Networks with Fast Adversarial Calibration

We introduce a new training strategy combining an entropy-encouraging loss term with anadversarial calibration loss term . This results in well-calibrated and technically trustworthy predictions for a wide range of domains drifts . We observe that our approach substantially outperforms existing state-of-the-artapproaches, yielding well-Calibrated predictions under domain drift .…

Reinforcement Learning based Multi Robot Classification via Scalable Communication Structure

In the multi-robot collaboration domain, training with Reinforcement Learning(RL) can become intractable . Performance starts to deteriorate drastically as the number of robots increases . In this work, we proposed a scalable communication structure . The proposed architecture achieves acomparable classification accuracy with the centralized methods, maintains highperformance with various numbers of robots without additional training cost, and robust to hacking and loss of the robots in the network .…

A Game Theoretic Perspective into the Coexistence of WiFi and NR U at the 6GHz Unlicensed Bands

We study behaviour of WiFi and 5G cellular networks as they exploit therecently unlocked 6 GHz spectrum for unlicensed access . We propose a framework where the portions of cellularand WiFi networks are grouped together to form entities . The action of an entity corresponds to the fraction ofits network elements (WiFi access point and cellular base stations) operatingin the 6 GHz band .…

Visual Speech Enhancement Without A Real Visual Stream

Current state-of-the-art methods use only the audiostream and are limited in their performance in a wide range of real-world noises . We propose a new paradigm for speech enhancement by exploiting recent breakthroughs in speech-driven lip synthesis . The intelligibility of the speech enhanced by our pseudo-lip approach is comparable to the case of using real lips.…

Probabilistic Dependency Graphs

Probabilistic Dependency Graphs (PDGs) can capture inconsistent beliefs in a natural way . PDGs are more modular than Bayesian Networks (BNs) in that they make it easier to incorporate new information and restructure the representation . We provide threesemantics for PDGs, each of which can be derived from a scoring function (onjoint distributions over the variables in the network) that can be viewed as representing a distribution’s incompatibility with the PDG .…

yNet a multi input convolutional network for ultra fast simulation of field evolvement

The proposed yNet is applied to the simulation ofporosity evolution in selective lasering sintering (SLS) Upon testing, yNet can simulate nearly identical porosity evolution and development to the physics-based model, with a 99.13% morphological similarity for various SLSconditions . yNet may have a transformative impact bydemocratizing the capability of ultra-fast and extreme-scale simulation ofvarious field evolvements.…

Multi Decoder Attention Model with Embedding Glimpse for Solving Vehicle Routing Problems

We present a novel deep reinforcement learning method to learn constructionheuristics for vehicle routing problems . In specific, we propose aMulti-Decoder Attention Model (MDAM) to train multiple diverse policies, whicheffectively increases the chance of finding good solutions . Extensive experiments on six different routing problems show thatour method significantly outperforms the state-of-the-art deep learning basedmodels .…

Minimax Strikes Back

Deep Reinforcement Learning (DRL) reaches a superhuman level of play in many complete information games . The state of the art search algorithm used incombination with DRL is Monte Carlo Tree Search (MCTS) We take another approach to DRL using a Minimax algorithm instead of MCTS and learning only the evaluation of states, not the policy .…

On Emergent Systematic Generalisation and Compositionality in Visual Referential Games with Straight Through Gumbel Softmax Estimator

The drivers of compositionality in artificial languages that emerge when two agents play a non-visual referential game has been previously investigated using approaches based on the REINFORCE algorithm and the (Neural)Iterated Learning Model . Compositionality and the generalisation abilities of theemergent languages are assessed using topographic similarity and zero-shotcompositional tests .…

Achieving Reliable Causal Inference with Data Mined Variables A Random Forest Approach to the Measurement Error Problem

Combining machine learning with econometric analysis is becoming increasingly popular . We propose employing random forest notjust for prediction, but also for generating instrumental variables to addressthe measurement error embedded in the prediction . We design a data-driven procedure to select tuples of individual trees from a random forest, in which one tree serves as the endogenous covariate and the other trees serve as its instruments .…

Uncertainty Aware Label Refinement for Sequence Labeling

Conditional random fields (CRF) for label decoding has become ubiquitous in sequence labeling tasks . However, the local label dependencies and inefficientViterbi decoding have always been a problem to be solved . In this work, we introduce a novel two-stage label decoding framework to model long-term labeldependencies, while being much more computationally efficient .…

Enabling Micro payments on IoT Devices using Bitcoin Lightning Network

LightningNetwork (LN) works on top of Bitcoin by leveraging off-chain transactions . The idea is to involve the IoT device in LN operations with its digital signature . LN’s Bitcoin transactions are revised toincorporate 3-of-3 signatures . We evaluated the proposed protocol by implementing it on a Raspberry Pifor a toll payment scenario and demonstrated its feasibility and security.…

Confused Modulo Projection based Somewhat Homomorphic Encryption Cryptosystem Library and Applications on Secure Smart Cities

With the development of cloud computing, the storage and processing of massive visual media data has gradually transferred to the cloud server . We propose a single-server version of somewhathomomorphic encryption cryptosystem based on confused modulo projection theoremnamed CMP-SWHE . On the clientside, the original data is encrypted by amplification, randomization, and setting confusing redundancy .…

FraCaS Temporal Analysis

In this paper, we propose an implementation of temporal semantics for inference problems . This implementation translates syntax trees tological formulas, suitable for consumption by Coq proof assistant . Wesupport several phenomena including: temporal references, temporal adverbs,aspectual classes and progressives .…

Hedge Connectivity without Hedge Overlaps

Connectivity is a central notion of graph theory and plays an important role in graph algorithm design and applications . With emerging new applications innetworks, a new type of graph connectivity problem has been getting moreattention . In this paper, we consider the model of hedgegraphs without hedge overlaps, where edges are partitioned into subsets calledhedges that fail together .…

On the Power of Localized Perceptron for Label Optimal Learning of Halfspaces with Adversarial Noise

We study active learning of homogeneous $s$-sparse halfspaces in $mathbb{R}^d$ with adversarial noise . Our main contribution is a state-of-the-art online active learning algorithm that achieves near-optimal attribute efficiency, label and sample complexity under mild distributional assumptions . We believethat our algorithmic design and analysis are of independent interest, and mayshed light on learning halfsaces with broader noise models .…

TOPCAT Visualisation over the Web

TOPCAT and STILTS offer visual exploration of locally stored tables containing millions of rows or more . Remote lightweight HTML/JavaScript clients can configure and interact with plots based on that data . The interaction can includepan/zoom/rotate navigation, identifying individual points, and potentially subset selection .…

Random pattern and frequency generation using a photonic reservoir computer with output feedback

Reservoir computing is a bio-inspired computing paradigm for processing timedindependent signals . The performance of its analogue implementations matches other digital algorithms on a series of benchmark tasks . Their potential can befurther increased by feeding the output signal back into the reservoir, which would allow to apply the algorithm to time series generation .…

Self Supervision based Task Specific Image Collection Summarization

Successful applications of deep learning (DL) requires large amount ofannotated data . This often restricts the benefits of employing DL to businesses and individuals with large budgets for data-collection and computation . Summarization offers a possible solution by creating much smallerrepresentative datasets that can allow real-time deep learning and analysis of big data and thus democratize use of DL .…

A unified implementation of automata and expression structures and of the associated algorithms using enriched categories

In this document, we propose a description, via a Haskell implementation, of a generalization of the notion of regular expression allowing us to group the definitions and the methods of (tree or word) automata constructions over one generic structure . The Haskell implementation and the algebraicdefinition of the generic automaton structure are based on the following ideas: – enriched categories, enriched functors, enriched monads, etc.…

Incentive based Decentralized Routing for Connected and Autonomous Vehicles using Information Propagation

Routing strategies under the aegis of dynamic traffic assignment have been proposed in the literature to optimize system performance . This study proposes anincentive-based decentralized routing strategy to nudge the network performance closer to the system optimum for the context where all vehicles are connected and autonomous vehicles (CAVs) The strategy consists of three stages.…

Modeling Silicon Photonic Neural Networks under Uncertainties

Silicon-photonic neural networks offer substantial improvements incomputing speed and energy efficiency compared to their digital electronic counterparts . However, the energy efficiency and accuracy of SPNNs are highly impacted by uncertainties that arise from fabrication-process and thermalvariations . In an SPNN with two hidden layers and 1374 tunable-thermal-phase shifters, random uncertainties can lead to a catastrophic 70% accuracy loss of accuracy loss in a SPNN .…

Overcoming Language Priors with Self supervised Learning for Visual Question Answering

Most Visual Question Answering (VQA) models suffer from the language priorproblem, which is caused by inherent data biases . VQA models tend to answer questions (e.g., what color is the banana?) based on the high-frequency answers . In this paper, we propose a self-supervised auxiliary task to utilize the balanced datato to assist the base VQ a model to overcome language priors .…

Smart Refrigerator using Internet of Things and Android

The kitchen is regarded as the central unit of the traditional as well as the traditional kitchen . The refrigerator is the pivotal of all that, and hence it plays animportant part in our regular lives . The idea of this project is to improvisethe normal refrigerator into a smart one by making it to place order for food items and to create an virtual interactive environment between it and the user .…

Visuo Locomotive Complexity as a Component of Parametric Systems for Architecture Design

A people-centred approach for designing large-scale built-up spacesnecessitates systematic anticipation of user’s embodied visuo-locomotive experience . In thiscontext, we develop a behaviour-based Visuo-Locomotive complexity model that functions as a key correlate of cognitive performance vis-a-vis internalnavigation . We present examples based on anempirical study in two healthcare buildings, and showcase the manner in which adynamic and interactive parametric (complexity) model can promote behavioural decision-making throughout the design process to maintain desired levels of visuospatial complexity as part of a navigation or wayfinding experience .…

Program Analysis an Appetizer

This book is an introduction to program analysis that is meant to beconsiderably more elementary than our advanced book Principles of ProgramAnalysis . Rather than using flow charts as the model of programs, the book follows our introductory book Formal Methods an Appetizer using program graphs .…

Network Distributed Video Coding

Network-distributed videocoding (NDVC) was proposed within the Moving Picture Experts Group (MPEG) Theaim of NDVC is to reduce the storage cost compared to simulcast, while retaining a smaller computing cost . NDVC offers a third option for video providers to deliver their contents to their clients .…