Pseudo supervised Deep Subspace Clustering

Auto-Encoder (AE)-based deep subspace clustering (DSC) methods have achievedimpressive performance due to the powerful representation extracted using deepneural networks . However, self-reconstruction loss of an AE ignores rich useful relation information and might lead to indiscriminative representation . Pseudo-graphs and pseudo-labels, which allow benefiting from uncertain knowledge acquired during network training, are further employed to supervise similarity learning .…

First order natural deduction in Agda

Agda is a dependently-typed functional programming language . It is based on an extension of intuitionistic Martin-L\”of type theory . We implement first order natural deduction in Agda . We use Agda’s type checker to verify the correctness of natural deduction proves properties of natural deduction .…

Risk Aware Lane Selection on Highway with Dynamic Obstacles

When lane-changing is discretionary, it is advised not to change lanes unless highly beneficial, e.g., reducing traveltime significantly or securing higher safety . We propose a real-time lane-selection algorithm with careful cost considerations and with modularity in design . The algorithm is asearch-based optimization method that evaluates uncertain dynamic positions of other vehicles under a continuous time and space domain .…

Generative Landmarks

Most sparse landmark detectionmethods rely on laborious, manually labelled landmarks . High-quality landmarks with personalization is often hard to achieve . We propose a general purpose approach to detect landmarks with improvedtemporal consistency . We capture two setsof unpaired marked (with paint) and unmarked videos .…

Learning What To Do by Simulating the Past

Recent work proposed that agents have access to a source of information that is effectively free: in any environment that humans have acted in, the state will already be optimized for human preferences . Such learning is possible in principle, but requires simulating all possible past trajectories that could have led to the observed state .…

Towards End to End Neural Face Authentication in the Wild Quantifying and Compensating for Directional Lighting Effects

The recent availability of low-power neural accelerator hardware, combined with improvements in end-to-end neural facial recognition algorithms provides,enabling technology for on-device facial authentication . The present researchwork examines the effects of directional lighting on a State-of-Art(SoA) neuralface recognizer . Top lighting and its variants (top-left, top-right) arefound to have minimal effect on accuracy .…

BR NS an Archive less Approach to Novelty Search

Novelty Search traditionallyrelies on k-nearest neighbours search and an archive of previously visited behavior descriptors which are assumed to live in a Euclidean space . The performance of NoveltySearch depends on a number of algorithmic choices and hyperparameters, such as the number of elements to add or remove elements to the archive .…

Spatial Imagination With Semantic Cognition for Mobile Robots

The imagination of the surrounding environment based on experience and semantic cognition has great potential to extend the limited observations and provide more information for mapping, collision avoidance, and path planning . We demonstrate that the algorithm can perform imagination for unseen parts of the object, by recalling the images and experience .…

Re designing cities with conditional adversarial networks

This paper introduces a conditional generative adversarial network to redesign a street-level image of urban scenes by generating an urban intervention policy, an attention map that localises where intervention is needed . The trained model shows strong performance in re-modellingcities, outperforming existing methods that apply image-to-image translation in other domains that is computed in a single GPU .…

A Reinforcement Learning Environment For Job Shop Scheduling

Finding schedules often intractable and cannot be achieved by Combinatorial OptimizationProblem (COP) methods within a given time limit . Recent advances of DeepReinforcement Learning (DRL) in learning complex behavior enable new COPapplication possibilities . We demonstrate that our approach significantlyoutperforms existing DRL methods on classic benchmark instances, coming closeto state-of-the-art COP approaches .…

Uncertainty aware Remaining Useful Life predictor

Remaining Useful Life (RUL) estimation is the problem of inferring how long an industrial asset can be expected to operate within its defined specifications . Machine Learning (ML) algorithms are natural candidates totackle the challenges involved in the design of intelligent maintenancesystems .…

Support Target Protocol for Meta Learning

The support/query (S/Q) training protocol is widely used in meta-learning . S/Q protocol trains a task-specific model on S and then evaluates it on Q tooptimize the meta-model using query loss, which depends on size and quality of Q . The new S/T protocol offers a more accurate evaluation since it does not rely on possibly biased and noisy query instances .…

Multimodal Fusion of EMG and Vision for Human Grasp Intent Inference in Prosthetic Hand Control

For lower arm amputees, robotic prosthetic hands offer the promise to regain the capability to perform fine object manipulation in activities of dailyliving . Current control methods based on physiological signals such as EEG and EMG are prone to poor inference outcomes due to motion artifacts, variabilityof skin electrode junction impedance over time, muscle fatigue, and other factors .…

Connecting Deep Reinforcement Learning based Obstacle Avoidance with Conventional Global Planners using Waypoint Generators

Deep Reinforcement Learning has emerged as an efficient dynamic obstacleavoidance method in highly dynamic environments . It has the potential toreplace overly conservative or inefficient navigation approaches . However, the integration of the method into existing navigation systems is still an open frontier due to the myopic nature of Deep Reinforcer Learning .…

Just Label What You Need Fine Grained Active Selection for Perception and Prediction through Partially Labeled Scenes

Self-driving vehicles must perceive and predict the future positions of nearby actors in order to avoid collisions and drive safely . Active learning techniques leverage the state of the current model to iteratively select examples for labeling . Wethus introduce generalizations that ensure that our approach is both cost-awareand allows for fine-grained selection of examples through partially labeled scenes .…

Software Hardware Co design for Multi modal Multi task Learning in Autonomous Systems

Optimizing quality of result (QoR) and the quality of service (QoS) of AI-empowered autonomous systems simultaneously is very challenging . Autonomous systems essentially require multi-modalmulti-task (MMMT) learning which must be aware of hardware performance and implementation strategies . We advocate for further explorations of MMMT in autonomous systems and software/hardware co-design solutions .…

Flavored Tacotron Conditional Learning for Prosodic linguistic Features

Yasuda et. al. show that Tacotron-2’s encoder doesn’t fully represent prosodicfeatures (e.g. syllable stress in English) from characters, and result in flatfundamental frequency variations . They propose a novel carefully designed strategy for conditioningTacotron 2 on two fundamental prosodic features in English — stress syllableand pitch accent — that help achieve more natural prosody .…

Unitary Subgroup Testing

We present a novel structural property of Clifford unitaries . Namely, that their (normalized) trace is bounded by $1/\sqrt{2$ in absolute value . We show a similar property for the $q$-aryCliffords . This allows us to analyze a simple single-query identity test under the Clifford promise and show that it has (at least) constant soundness .…

On Mixed Iterated Revisions

The ten operators considered in this article are shownto be all reducible to three: lexicographic revision, refinement and severewithdrawal . The complexity of mixed sequences of belief change operators is alsoanalyzed . Most of them require only a polynomial number of calls to asatisfiability checker, some are even easier.…

Voluntary safety commitments provide an escape from over regulation in AI development

With the introduction of Artificial Intelligence (AI) and relatedtechnologies in our daily lives, fear and anxiety about their misuse as well asthe hidden biases in their creation have led to a demand for regulation . Yet blindly regulating an innovation process that is notwell understood, may stifle this process and reduce benefits that society may gain from the generated technology, even under the best intentions .…

Post Hoc Domain Adaptation via Guided Data Homogenization

Addressing shifts in data distributions is an important prerequisite for the deployment of deep learning models to real-world settings . We demonstrate the potential of data homogenization through experiments on the CIFAR-10 and MNIST data sets . This approach makes use of information about the training data contained implicitly in the deep learning model to learn a domain transfer function .…

The virtual element method for the coupled system of magneto hydrodynamics

Virtual Element Method allows us to construct noveldiscretizations for simulating realistic phenomenon in magneto-hydrodynamics . We show that this VEM approximation will yield divergence freediscrete magnetic fields, an important property in any simulation in MHD . We present a model for magneticreconnection in a mesh that includes a series of hanging nodes, which we use tocalibrate the resolution of the method .…

Advances in Metric Ramsey Theory and its Applications

Metric Ramsey theory is concerned with finding large well-structured subsetsof more complex metric spaces . For finite metric spaces this problem was first studied by Bourgain, Figiel and Milman . In this paper we provide deterministicconstructions for this problem via a novel notion of \emph{metric Ramseydecomposition}.…

Sound Probabilistic Inference via Guide Types

Probabilistic programming languages aim to describe and automate Bayesianmodeling and inference . Modern languages support programmable inference, which allows users to customize inference algorithms by incorporating guide programs . For Bayesian inference to be sound, guide programs must be compatible with model programs .…