Showing papers by "Andrei Rusu published in 2019"

Continual Unsupervised Representation Learning.

Abstract: Continual learning aims to improve the ability of modern learning systems to deal with non-stationary distributions, typically by attempting to learn a series of tasks sequentially. Prior art in the field has largely considered supervised or reinforcement learning tasks, and often assumes full knowledge of task labels and boundaries. In this work, we propose an approach (CURL) to tackle a more general problem that we will refer to as unsupervised continual learning. The focus is on learning representations without any knowledge about task identity, and we explore scenarios when there are abrupt changes between tasks, smooth transitions from one task to another, or even when the data is shuffled. The proposed approach performs task inference directly within the model, is able to dynamically expand to capture new concepts over its lifetime, and incorporates additional rehearsal-based techniques to deal with catastrophic forgetting. We demonstrate the efficacy of CURL in an unsupervised learning setting with MNIST and Omniglot, where the lack of labels ensures no information is leaked about the task. Further, we demonstrate strong performance compared to prior art in an i.i.d setting, or when adapting the technique to supervised tasks such as incremental class learning.

Meta-Learning with Warped Gradient Descent

Abstract: Learning an efficient update rule from data that promotes rapid learning of new tasks from the same distribution remains an open problem in meta-learning. Typically, previous works have approached this issue either by attempting to train a neural network that directly produces updates or by attempting to learn better initialisations or scaling factors for a gradient-based update rule. Both of these approaches pose challenges. On one hand, directly producing an update forgoes a useful inductive bias and can easily lead to non-converging behaviour. On the other hand, approaches that try to control a gradient-based update rule typically resort to computing gradients through the learning process to obtain their meta-gradients, leading to methods that can not scale beyond few-shot task adaptation. In this work, we propose Warped Gradient Descent (WarpGrad), a method that intersects these approaches to mitigate their limitations. WarpGrad meta-learns an efficiently parameterised preconditioning matrix that facilitates gradient descent across the task distribution. Preconditioning arises by interleaving non-linear layers, referred to as warp-layers, between the layers of a task-learner. Warp-layers are meta-learned without backpropagating through the task training process in a manner similar to methods that learn to directly produce updates. WarpGrad is computationally efficient, easy to implement, and can scale to arbitrarily large meta-learning problems. We provide a geometrical interpretation of the approach and evaluate its effectiveness in a variety of settings, including few-shot, standard supervised, continual and reinforcement learning.

Effectiveness of job crafting interventions: a meta-analysis and utility analysis

Abstract: Job crafting (JC) is a form of bottom-up job design with a high potential for increasing work engagement and performance. For this reason, researchers have proposed interventions to stimulate JC in...

Task Agnostic Continual Learning via Meta Learning

Abstract: While neural networks are powerful function approximators, they suffer from catastrophic forgetting when the data distribution is not stationary. One particular formalism that studies learning under non-stationary distribution is provided by continual learning, where the non-stationarity is imposed by a sequence of distinct tasks. Most methods in this space assume, however, the knowledge of task boundaries, and focus on alleviating catastrophic forgetting. In this work, we depart from this view and move the focus towards faster remembering -- i.e measuring how quickly the network recovers performance rather than measuring the network's performance without any adaptation. We argue that in many settings this can be more effective and that it opens the door to combining meta-learning and continual learning techniques, leveraging their complementary advantages. We propose a framework specific for the scenario where no information about task boundaries or task identity is given. It relies on a separation of concerns into what task is being solved and how the task should be solved. This framework is implemented by differentiating task specific parameters from task agnostic parameters, where the latter are optimized in a continual meta learning fashion, without access to multiple tasks at the same time. We showcase this framework in a supervised learning scenario and discuss the implication of the proposed formalism.

Romanian Version of the Internet Addiction Test: Psychometric Properties and Cross-Gender Invariance

Abstract: Nowadays, excessive use of the Internet and addiction to it became constant concerns among researchers. The proliferation of this research stream came along with a constant need for valid operationalizations. Our study aimed to test the psychometric properties of the Internet Addiction Test (IAT) on a Romanian sample (n = 421 students). Exploratory factor analysis revealed a two-factor solution: Mental Disorder (α = .78); Time Management Disorder (α = .79). Also, the overall IAT had good reliability (α = .86). Confirmatory factor analysis supported the factorial structure and through multigroup comparisons revealed that the instrument is invariant across genders. Further proofs in support of the IAT’s validity were given by the significant associations with social anxiety, social support, and dissociation. Moreover, each scale and also the overall IAT score correlated positively with participants’ self-reported time spent online (both during a week’s working days and during weekends), and the time spent on the Facebook messenger smartphone application. The study showed that the Romanian version of the IAT has promising psychometric properties and can be successfully used in research regarding excessive Internet use.

Continual Unsupervised Representation Learning

Abstract: Continual learning aims to improve the ability of modern learning systems to deal with non-stationary distributions, typically by attempting to learn a series of tasks sequentially. Prior art in the field has largely considered supervised or reinforcement learning tasks, and often assumes full knowledge of task labels and boundaries. In this work, we propose an approach (CURL) to tackle a more general problem that we will refer to as unsupervised continual learning. The focus is on learning representations without any knowledge about task identity, and we explore scenarios when there are abrupt changes between tasks, smooth transitions from one task to another, or even when the data is shuffled. The proposed approach performs task inference directly within the model, is able to dynamically expand to capture new concepts over its lifetime, and incorporates additional rehearsal-based techniques to deal with catastrophic forgetting. We demonstrate the efficacy of CURL in an unsupervised learning setting with MNIST and Omniglot, where the lack of labels ensures no information is leaked about the task. Further, we demonstrate strong performance compared to prior art in an i.i.d setting, or when adapting the technique to supervised tasks such as incremental class learning.

Using Adaptive Transmit Power in Wireless Indoor Air Quality Monitoring

Abstract: In the past few years, the general interest in connected home, smart devices and appliances has risen significantly. New concepts like the Internet of Things have inspired new applications and research areas for sensors and gathering data from them. One important application that makes use of these new technologies is monitoring indoor air quality, which according to Directive 2010/31/EU of the European Parliament and of the Council, can “reduce mortality, morbidity, and health care costs”. In this context, this paper proposes a solution for wireless monitoring of indoor air quality parameters including temperature, humidity, CO2 and TVOC levels using a low-power ARM Cortex-M microcontroller, an IEEE 802.15.4 sub-GHz transceiver and temperature, humidity and gas sensors and observing the changes in overall power consumption by implementing an adaptive transmit power technique.

Study on the mapping of research data in the republic of moldova in the context of open science

Abstract: The Open Science concept represents a new approach to the way in which scientific research based on cooperation and new ways of knowledge dissemination is carried out and organized, using new digital technologies, new tools for collaboration, and R&D infrastructure to ensure open access to research data. This study uses data collected in May - July 2018 within a survey that aimed at investigating the scientific data ecosystem in the Republic of Moldova. Findings show that, although there are some concerns about the loss of property rights and copyright infringement in case of sharing and open access to research data, Moldovan academia is ready to provide access to research data. The research has highlighted that a new challenge is needed to solve scientific data issues by creating new type of infrastructure to ensure data retention, broad access to research results for the purpose of their dissemination and use, and creating new research opportunities based on research data.

Channel Hopping in Wireless Process Control

Abstract: Most of the applications in the industrial area use wireless for process monitoring where traditional wired sensors cannot be installed like, for example, hazardous or explosive environments. The use of wireless communication for process control has often been seen as troublesome with a lack of robustness and determinism, range problems and interference. Possible solutions include redundancy for robustness, channel hopping for interference and a higher transmit power for range problems. This paper aims to cover the interference issue by proposing a possible solution with channel hopping patterns, implemented on a low-power Arm based micro-controller paired with an IEEE 802.15.4 sub-GHz transceiver.