Does Standard Backpropagation Forget Less Catastrophically Than Adam

Catastrophic forgetting remains a severe hindrance to the broad application of artificial neural networks, however, it continues to be a poorlyunderstood phenomenon . We argue that it is still unclear how exactly the phenomenon should be quantified . We recommend inter-task forgetting insupervised learning must be measured with both retention and relearning metrics co-concurrently .…

CQNet Complex Input Quantized Neural Network designed for Massive MIMO CSI Feedback

The Massive Multiple Input Multiple Output (MIMO) system is a core technology of the next generation communication . Traditional compressive sensing based CSI feedback hasbecome a bottleneck problem that is limited in piratical . CQNet outperforms the state-of-the-art method with less computational overhead by achieving an average performanceimprovement of 8.07% in both outdoor and indoor scenarios .…

On Ray Shooting for Triangles in 3 Space and Related Problems

We consider several problems that involve lines in three dimensions, and present improved algorithms for solving them . Our approach is based on the polynomial partitioning technique . For example, our ray-shooting algorithm processes a set of $n$ triangles in $R^3$ into a data structure for answering ray shooting queries amid the giventriangles .…

End to End Egospheric Spatial Memory

Egospheric SpatialMemory encodes the memory in an ego-sphere around the agent, enabling expressive 3D representations . ESM can be trained end-to-end via imitation or reinforcement learning, and improves both trainingefficiency and final performance against other memory baselines . The explicit egocentric geometry also enables us to seamlessly combine the learned controller with other non-learnedmodalities, such as local obstacle avoidance.…

Data Driven Retrospective Cost Adaptive Control for Flight Control Application

The paper investigates the ability of RLS-VRF to provide themodeling information needed for the target model, especially nonminimum-phase(NMP) zeros . DDRCAC is applied to single-input, single-output (SISO) and multiple-input (MIMO) numerical examples with unknown NMPzeros, as well as several flight control problems, namely, unknown transitionfrom minimum-phase to NMP lateral dynamics, flexible modes, flutter, and nonlinear planar missile dynamics .…

The corruptive force of AI generated advice

Artificial Intelligence (AI) is increasingly becoming a trusted advisor in people’s lives . A new concern arises if AI persuades people to break ethical rules for profit . Employing a large-scale behavioural experiment (N = 1,572), we test whether AI-generated advice can corrupt people .…

Machine Learning Model Development from a Software Engineering Perspective A Systematic Literature Review

The problems regarding Machine LearningDevelopment involves the fact that such professionals do not realize that they perform ad-hoc practices that could be improved by the adoption of the Software Engineering Development Lifecycle . Ofcourse, since machine learning systems are different from traditional Softwaresystems, some differences in their respective development processes are to beexpected .…

Cooperation and Reputation Dynamics with Reinforcement Learning

Creating incentives for cooperation is a challenge in natural and artificial systems . One potential answer is reputation, whereby agents trade the immediate cost of cooperation for the future benefits of having a good reputation . We use a simple model ofreinforcement learning to show that reputation mechanisms generate twocoordination problems .…

Secure UCB Saving Stochastic Bandits from Poisoning Attacks via Limited Data Verification

This paper studies bandit algorithms under data poisoning attacks in abounded reward setting . We consider a strong attacker model in which theattacker can observe both the selected actions and their corresponding rewards . We show that there exists an $O(\logT)$ regret bandit algorithm, specifically the classical UCB, that requires $Omega(T) amount of contamination to suffer regret .…

Multimodal Mobility Systems Joint Optimization of Transit Network Design and Pricing

The performance of multimodal mobility systems relies on the seamlessintegration of conventional mass transit services and the advent ofMobility-on-Demand (MoD) services . A primal-dual approach, inspired by the market literature, yields a compact mixed integer linear programming (MILP)formulation . We provide atractable solution approach through a decomposition scheme and approximational algorithm that accelerates the computation and enables optimization of large-scale problem instances .…

Efficient solvers for shallow water Saint Venant equations and debris transportation deposition models

This research is aimed at achieving an efficient digital infrastructure forevaluating risks and damages caused by tsunami flooding . Rather than using complex multiphase debris models, werather use an empirical transportation and deposition model that takes into account the interaction with the main water flow, friction/contact with the ground but also debris interaction .…

DOBF A Deobfuscation Pre Training Objective for Programming Languages

Recent advances in self-supervised learning have dramatically improved the state of the art on a wide variety of tasks . However, research in languagemodel pre-training has mostly focused on natural languages . In this paper, we introduce a new objective, DOBF, that leverages the structural aspect of programming languages and pre-trains a model to recover the original version of obfuscated source code .…

Fast End to End Speech Recognition via Non Autoregressive Models and Cross Modal Knowledge Transferring from BERT

Attention-based encoder-decoder (AED) models have achieved promising performance in speech recognition . However, because the decoder predicts texttokens in an autoregressive manner, it is difficult for an AED model to predict all tokens in parallel . We propose a non-autoregressive speech recognition model called LASO(Listen Attentively, and Spell Once) The model consists of an encoder, adecoder, and a position dependent summarizer (PDS) The three modules are based on basic attention blocks.…

MAPGN MAsked Pointer Generator network for sequence to sequence pre training

Spoken-text normalization that converts spoken-style text into style normalized text is becoming an important technology for improving subsequent processing such as machine translation andsummarization . MAsked Pointer-Generator Network(MAPGN) is a novel self-supervised learning method . MAPGN is more effective for pointer-generatornetworks than the conventional methods in twospoken text normalization tasks .…

AI Ethics Needs Good Data

In this chapter we argue that discourses on AI must transcend the language of’ethics’ and engage with power and political economy in order to constitute’Good Data’ In particular, we must move beyond the depoliticised language of ‘ethics’, currently deployed (Wagner 2018) in determining whether AI is ‘good’ given the limitations of ethics as a frame through which AI issues can be viewed .…

OntoZSL Ontology enhanced Zero shot Learning

Zero-shot Learning (ZSL) aims to predict for those classes that havenever appeared in the training data . The key of implementing ZSL is to leverage the prior knowledge of classes which buildsthe semantic relationship between classes and enables the transfer of the learned models (e.g.,…

A Simple Deep Equilibrium Model Converges to Global Optima with Weight Tying

A deep equilibrium linear model is implicitly defined through an equilibriumpoint of an infinite sequence of computation . It avoids any explicitcomputation of the infinite sequence by finding an equilibrium point directly . Despite non-convexity, convergence to globaloptimum at a linear rate is guaranteed without any assumption on the width of the models, allowing the width to be smaller than the output dimension and thenumber of data points .…

Weak Adaptation Learning Addressing Cross domain Data Insufficiency with Weak Annotator

Data quantity and quality are crucial factors for data-driven learningmethods . In some target problem domains, there are not many data samples available, which could significantly hinder the learning process . We propose a weak adaptationlearning (WAL) approach that leverages unlabeled data from a similar sourcedomain, a low-cost weak annotator that produces labels based on task-specificheuristics, labeling rules, or other methods (albeit with inaccuracy), and asmall amount of labeled data in the target domain .…

Consistency based Merging of Variability Models

Globally operating enterprises selling large and complex products and services often have to deal with situations where variability models are locally developed to take into account the requirements of local markets . Forexample, cars sold on the U.S. market are represented by variability models insome or many aspects different from European ones .…

Learning from Demonstrations using Signal Temporal Logic

Learning-from-demonstrations is an emerging paradigm to obtain effectiverobot control policies for complex tasks via reinforcement learning without the need to explicitly design reward functions . We use Signal Temporal Logic to evaluate and rank the quality of demonstrations . We validate our approach through experiments on therete-world and OpenAI Gym environments .…

How RL Agents Behave When Their Actions Are Modified

Reinforcement learning in complex environments may require supervision toprevent the agent from attempting dangerous actions . We present the Modified-Action MarkovDecision Process, an extension of the MDP model that allows actions to differ from the policy . We analyze the asymptotic behaviours of common reinforcementlearning algorithms in this setting and show that they adapt in different ways: Some completely ignore modifications while others go to various lengths intrying to avoid action modifications that decrease reward .…

Cross modal Adversarial Reprogramming

Recent works onadversarial reprogramming have shown that it is possible to repurpose neuralnetworks for alternate tasks without modifying the network architecture or parameters . In this work, we broaden the scope of adversarialreprogramming beyond the data modality of the original task .…

Prompt Programming for Large Language Models Beyond the Few Shot Paradigm

In this work, we discuss methods of promptprogramming, emphasizing the usefulness of considering prompts through the lensof natural language . We explore techniques for exploiting the capacity of narrative and cultural anchors to encode nuanced intentions . We suggest that the function of few-shot examples in these cases is better described as locating an already learned task rather than meta-learning .…

CHARET Character centered Approach to Emotion Tracking in Stories

Autonomous agents that can engage in social interactions with a human is the ultimate goal of a myriad of applications . Wepropose a characterrole-labelling approach to emotion tracking that accounts for the semantics of emotions . We show that by identifying characters and objects of events andconsidering the emotional state of the characters, we can achieve betterperformance in this task .…

Data driven Analysis for Understanding Team Sports Behaviors

Real-world biological multi-agent behaviors such as team sports are often largely unknown due to their inherently higher-order interactions,cognition, and body dynamics . Estimation of the rules from data provides an effective way for the analysis of such behaviors . Survey focuses on data-driven analysis for quantitativeunderstanding of team sports behaviors .…

Leveraging Acoustic and Linguistic Embeddings from Pretrained speech and language Models for Intent Classification

An intent classification system is usually implemented as a pipeline process, with aspeech recognition module followed by text processing that classifies theintents . Such systems don’t take advantage of relevant linguistic information, and suffer from limited training data . In this work, we propose a novel intent classification framework that uses acoustic features extracted from a speech recognition system and linguistic features learned from a pretrained language model .…