Market Value of Differentially Private Smart Meter Data

This paper proposes a framework to investigate the value of sharingprivacy-protected smart meter data between domestic consumers and load servingentities . The framework consists of a discounted differential privacy model toensure individuals cannot be identified from aggregated data . It also includes a ANN-basedshort-term load forecasting to quantify the impact of data availability and privacy protection on the forecasting error and an optimal procurement day-ahead and balancing markets to assess the market value of the privacy-utility trade-off .…

Development of digitally obtainable 10 year risk scores for depression and anxiety in the general population

The burden of depression and anxiety in the world is rising . Identification of individuals at increased risk of developing these conditions would help totarget them for prevention and ultimately reduce the healthcare burden . Wedeveloped a 10-year predictive algorithm using thefull cohort of over 400,000 UK Biobank participants without pre-existingdepression or anxiety using digitally obtainable information from the UKBiobank .…

Gradient Matching for Domain Generalization

Machine learning systems typically assume that the distributions of training and test sets match closely . We propose an inter-domain gradient matching objective that targets domaingeneralization by maximizing the inner product between gradients from different domains . We demonstrate the efficacy of Fish on 6 datasets from the Wildsbenchmark, which captures distribution shift across a diverse range ofmodalities .…

Class Incremental Learning with Generative Classifiers

Most existing class-incremental learning methods storedata or use generative replay . Instead of directly learning the conditional distributionp(y|x), our proposal is to learn the joint distribution p(x,y), factorized asp(x|y) p(y), and to perform classification using Bayes’ rule . This simple approach performs very well on adiverse set of continual learning benchmarks, outperforminggenerative replayand other existing baselines that do not store data .…

An Initial Algebra Theorem Without Iteration

The Initial Algebra Theorem states that an endofunctor has an initial algebra provided it has a pre-fixed point . The proof crucially depends on transfinitely iterating the functor and shows that the (transfinite)initial-algebra chain stops . This is madepossible by using Pataraia’s .…

IIoT Enabled Health Monitoring for Integrated Heat Pump System Using Mixture Slow Feature Analysis

Heat pump (HP) system is widely deployed in modern buildings for heating use . Many HPs werepractically manufactured and installed many years ago, resulting in fewersensors available due to technology limitations and cost control at that time . We propose a hybrid scheme by integrating industrial Internet-of-Things (IIoT) and intelligent health monitoring algorithms to handle this challenge .…

Multiscale deep context modeling for lossless point cloud geometry compression

We propose a practical deep generative approach for lossless point cloudgeometry compression . MSVoxelDNN significantly reduces compression rate compared to the MPEG G-PCC codec . The implementation is available athttps://://://github.com/Weafre/MSVoxelsDNN . The current method speeds up encoding/decodingtimes significantly compared to previous Voxel DNN, while having averagerate saving of 17.5% over G-pCC .…

The principle of weight divergence facilitation for unsupervised pattern recognition in spiking neural networks

Parallels between the signal processing tasks and biological neurons lead to an understanding of the principles of self-organized optimization of input signals . We propose the addition to the well-knownSTDP synaptic plasticity rule to directs the weight modification towards the state associated with the maximal difference between the background noise and related signals .…

Approximation of fractional harmonic maps

This paper addresses the approximation of fractional harmonic maps . The compactness results imply the convergence of numerical approximations . Numerical examples on spin chain dynamics and point defects are presented to demonstrate the effectiveness of the proposed methods . We devise and analyze numericalmethods for the .…

Fluid beam interaction Capturing the effect of embedded slender bodies on global fluid flow and vice versa

This work addresses research questions arising from the application of geometrically exact beam theory in the context of fluid-structure interaction(FSI) We propose amixed-dimensional embedded finite element approach for the coupling of one-dimensional equations to a three-dimensional background fluid mesh . Here, the fluid is described by theincompressible isothermal Navier-Stokes equations for Newtonian fluids .…

Role Aware Modeling for N ary Relational Knowledge Bases

N-ary relational knowledge bases (KBs) represent knowledge with binary and beyond-binary relational facts . But existing approaches are often directly extended from binary relational KBs, while missing the important semantic property of role . We propose a Role-Aware Modeling Modeling, RAM, for facts in n-ary .…

On the Impact of Word Error Rate on Acoustic Linguistic Speech Emotion Recognition An Update for the Deep Learning Era

Text encodings from automatic speech recognition (ASR) transcripts and audiorepresentations have shown promise in speech emotion recognition (SER) eversince . Yet, it is challenging to explain the effect of each information stream on the SER systems . More clarification is required for analysing the impact of ASR’s word error rate (WER) on linguistic emotion recognition per seand in the context of fusion with acoustic information exploitation in the age of deep ASR systems .…

VideoGPT Video Generation using VQ VAE and Transformers

VideoGPT uses VQ-VAE that learns downsampled latent latent representations of a raw video by learning 3D convolutions and axial self-attention . A simple GPT-like architecture is then used to autoregressively model the discrete latents usingspatio-temporal position encodings . Despite the simplicity in formulation and ease of training, our architecture is able to generate samples competitive with state-of-the-art GAN models for video generation on the BAIR Robot dataset and generate high fidelity natural images from UCF-101 and Tumbler GIF Dataset .…

Autonomous Situational Awareness for UAS Swarms

This paper describes a technique for the autonomous mission planning of unmanned aerial system swarms . Given a swarm operating in a known area, acentral command system generates measurements from the swarm . If those measurements indicate changes to the mission situation such as target movement, swarm planning is updated to reflect the new situation .…

Optimal Design of Electric Micromobility Vehicles

This paper presents a modeling and optimization framework to design batteryelectric micromobility vehicles, minimizing their total cost of ownership . We identify a model of the electric powertrain of ane-scooter and an e-moped consisting of a battery, a single electric motor and atransmission .…

An Efficient Approach for Anomaly Detection in Traffic Videos

The proposed approach comprises apre-processing module that detects changes in the scene and removes thecorrupted frames . A backtracking anomaly detection algorithm computes asimilarity statistic and decides on the onset time of the anomaly . Experimental results on the Track 4 test set of the 2021 AI City Challenge show the efficacy of the proposed framework as we achieve an F1-score of 0.9157 along with 8.4027 rootmean square error (RMSE) and are ranked fourth in the competition .…

Review of end to end speech synthesis technology based on deep learning

Speechsynthesis technology helps users get the output of intelligent machine more easily and intuitively . Current research focus is the deep learning-basedend-to-end speech synthesis technology . It mainly consists of three modules: textfront-end, acoustic model, and vocoder . This paper reviews the research status of these three parts, and classifies and compares various methods according to their emphasis .…

RoFormer Enhanced Transformer with Rotary Position Embedding

Position encoding in transformer architecture provides supervision for dependency modeling between elements at different positions in the sequence . The proposed RoPE encodes absolute positionalinformation with rotation matrix and naturally incorporates explicit relativeposition dependency in self-attention formulation . RoPE comes with flexibility of being expand to any sequencelengths, decaying inter-token dependency with increasing relative distances .…

Variational Relational Point Completion Network

Existing point cloud completion methods tend to generate globalshape skeletons and hence lack fine local details . They mostlylearn a deterministic partial-to-complete mapping, but overlook structuralrelations in man-made objects . To tackle these challenges, this paper proposes a Variational Relational point Completion network with two appealing properties: 1) Probabilistic Modeling.…

CoDR Computation and Data Reuse Aware CNN Accelerator

Computation and Data Reuse is critical for the resource-limited ConvolutionalNeural Network (CNN) accelerators . This paper presents Universal ComputationReuse to exploit weight sparsity, repetition, and similarity simultaneously in a convolutional layer . Compared to two recent compressed CNN accelerators with the same area of 2.85 mm^2, CoDR decreases SRAM access by 5.08x and 7.99x .…

Permutation Invariant Variational Autoencoder for Graph Level Representation Learning

Recently, there has been great success in applying deep neural networks ongraph structured data . But graph-level unsupervised learning has not received much attention yet . We propose a permutation-invariant variational autoencoder for graph structured data. Our proposed model indirectly learns to match the node ordering of input and output graph, without imposing a particular node ordering or performing expensive graph matching.…

Demystifying Regular Expression Bugs A comprehensive study on regular expression bug causes fixes and testing

An empirical study of 356 mergedregex-related pull request bugs from Apache, Mozilla, Facebook, and GoogleGitHub repositories . Correct regular expression behavior is the dominant root cause of regular expression bugs . The remaining root causes are incorrect API usage (9.3%)and other code issues that require regular expression changes in the fix(29.5%) Fixing regular expressions is nontrivial as it takes more time and more lines of code to fix them compared to general pull requests .…

Hierarchically Modeling Micro and Macro Behaviors via Multi Task Learning for Conversion Rate Prediction

Conversion Rate (\emph{CVR) prediction in modern industrial e-commerce platforms is becoming increasingly important . We propose a novel CVR prediction method by Hierarchically Modeling both Micro and Macro behaviors ($HM^3$) By employing multi-task learning andleveraging the abundant supervisory labels from micro and macro behaviors,$HM$ can be trained end-to-end and address the \emph {SSB} and $HM$ issues .…

Understanding Synonymous Referring Expressions via Contrastive Features

Referring expression comprehension aims to localize objects identified by natural language descriptions . Eachobject can be described by synonymous sentences with paraphrases, and suchvarieties in languages have critical impact on learning a comprehension model . We develop an end-to-end trainable framework to learn contrastive features on the image andobject instance levels .…

Grammatical Error Generation Based on Translated Fragments

We perform neural machine translation of sentence fragments in order to create large amounts of training data for English grammatical error correction . Our method aims at simulating mistakes made by second language learners . A model trained on data created using our proposed method is shownto outperform a baseline model on test data with a high proportion of errors .…

SelfReg Self supervised Contrastive Regularization for Domain Generalization

The proposed approach use only positive data pairs,thus it resolves various problems caused by negative pair sampling . The proposed method shows comparable performance to the conventional state-of-the-art alternatives . We propose a class-specific domain perturbation layer (CDPL), which makes itpossible to effectively apply mixup augmentation even when only positive datapairs are used .…

Efficient Retrieval Optimized Multi task Learning

Recently, there have been significant advances in neural methods for tackling knowledge-intensive tasks such as open domain question answering . Using our framework, we achieve comparable or better performance than recent methods on QA, while drastically reducing thenumber of parameters .…