Search

Human infant learning happens during exploration of the environment, by interaction with objects, and by listening to and repeating utterances casually, which is analogous to unsupervised learning. Only occasionally, a learning infant would receive a matching verbal description of an action it is committing, which is similar to supervised learning. Such a learning mechanism can be mimicked with deep learning. We model this weakly supervised learning paradigm using our Paired Gated Autoencoders (PGAE) model, which combines an action and a language autoencoder. After observing a performance drop when reducing the proportion of supervised training, we introduce the Paired Transformed Autoencoders (PTAE) model, using Transformer-based crossmodal attention. PTAE achieves significantly higher accuracy in language-to-action and action-to-language translations, particularly in realistic but difficult cases when only few supervised training samples are available. We also test whether the trained model behaves realistically with conflicting multimodal input. In accordance with the concept of incongruence in psychology, conflict deteriorates the model output. Conflicting action input has a more severe impact than conflicting language input, and more conflicting features lead to larger interference. PTAE can be trained on mostly unlabelled data where labeled data is scarce, and it behaves plausibly when tested with incongruent input., Comment: Published in: Applied Artificial Intelligence, 37:1, 2179167

We study a class of reinforcement learning problems where the reward signals for policy learning are generated by a discriminator that is dependent on and jointly optimized with the policy. This interdependence between the policy and the discriminator leads to an unstable learning process because reward signals from an immature discriminator are noisy and impede policy learning, and conversely, an under-optimized policy impedes discriminator learning. We call this learning setting \textit{Internally Rewarded Reinforcement Learning} (IRRL) as the reward is not provided directly by the environment but \textit{internally} by the discriminator. In this paper, we formally formulate IRRL and present a class of problems that belong to IRRL. We theoretically derive and empirically analyze the effect of the reward function in IRRL and based on these analyses propose the clipped linear reward function. Experimental results show that the proposed reward function can consistently stabilize the training process by reducing the impact of reward noise, which leads to faster convergence and higher performance compared with baselines in diverse tasks., Comment: Accepted at ICML 2023. Project webpage at https://ir-rl.github.io

Objective This study aims to understand borderline personality disorder (BPD) features by employing the Personality Psychopathology Five (PSY-5) scales from the Minnesota Multiphasic Personality Inventory-2 (MMPI-2).Methods A total of 156 psychiatric patients completed PSY-5 scales of MMPI-2 and Personality Assessment Inventory-Borderline Subscale (PAI-BOR). Pearson’s partial correlation analysis was conducted to control the impact of age and gender and to determine the relationship between PSY-5 scales and BOR. A hierarchical multiple regression analysis was implemented to examine whether PSY-5 scales predicted the BOR-total, and a path analysis was performed to determine whether PSY-5 scales predicted each PAI-BOR subscale.Results The BOR-total score had a significant correlation with all PSY-5 scores, even after controlling for age and gender. However, only aggressiveness (AGGR), disconstraint (DISC), negative emotionality/neuroticism (NEGE), and introversion/low positive emotionality (INTR), excluding psychoticism (PSYC), significantly predicted BOR-total. The path analysis indicates that PSYC did not predict any BOR subscale, while NEGE predicted all BOR subscales.Conclusion The study findings indicate that NEGE best reflects BPD features, while PSYC is far from the core domain that describes BPD. In addition, the influence of age should be considered when understanding BPD, since age predicted the BOR-total and two BOR subscales.

Spatial reasoning poses a particular challenge for intelligent agents and is at the same time a prerequisite for their successful interaction and communication in the physical world. One such reasoning task is to describe the position of a target object with respect to the intrinsic orientation of some reference object via relative directions. In this paper, we introduce GRiD-A-3D, a novel diagnostic visual question-answering (VQA) dataset based on abstract objects. Our dataset allows for a fine-grained analysis of end-to-end VQA models' capabilities to ground relative directions. At the same time, model training requires considerably fewer computational resources compared with existing datasets, yet yields a comparable or even higher performance. Along with the new dataset, we provide a thorough evaluation based on two widely known end-to-end VQA architectures trained on GRiD-A-3D. We demonstrate that within a few epochs, the subtasks required to reason over relative directions, such as recognizing and locating objects in a scene and estimating their intrinsic orientations, are learned in the order in which relative directions are intuitively processed.

Knowledge about space and time is necessary to solve problems in the physical world: An AI agent situated in the physical world and interacting with objects often needs to reason about positions of and relations between objects; and as soon as the agent plans its actions to solve a task, it needs to consider the temporal aspect (e.g., what actions to perform over time). Spatio-temporal knowledge, however, is required beyond interacting with the physical world, and is also often transferred to the abstract world of concepts through analogies and metaphors (e.g., "a threat that is hanging over our heads"). As spatial and temporal reasoning is ubiquitous, different attempts have been made to integrate this into AI systems. In the area of knowledge representation, spatial and temporal reasoning has been largely limited to modeling objects and relations and developing reasoning methods to verify statements about objects and relations. On the other hand, neural network researchers have tried to teach models to learn spatial relations from data with limited reasoning capabilities. Bridging the gap between these two approaches in a mutually beneficial way could allow us to tackle many complex real-world problems, such as natural language processing, visual question answering, and semantic image segmentation. In this chapter, we view this integration problem from the perspective of Neuro-Symbolic AI. Specifically, we propose a synergy between logical reasoning and machine learning that will be grounded on spatial and temporal knowledge. Describing some successful applications, remaining challenges, and evaluation datasets pertaining to this direction is the main topic of this contribution., Comment: Contribution to the book "A Compendium of Neuro-Symbolic Artificial Intelligence", which is to appear in the first half of 2023

Supplemental material, revision_appendix_1 for Clinical validation of screening test for at-risk drinking for young and middle-aged adults in an emergency department by Hosub Chung and Jae Hee Lee in Hong Kong Journal of Emergency Medicine

Additional file 1. General and clinical characteristics of study patients according to AKI stages

Everything I Love Is on the Table is a collection of poetry that centers on migration, history, abuse, and mental illness. The collection investigates the intersections of family, romantic love, and diasporic identity via domestic scenes meant to serve as a window into trauma. Ultimately, it is my attempt at remembering, archiving, and coming to terms with memories of suffering, which have been fragmented and distorted over the years, so that I may understand in a more profound manner the ways in which trauma has influenced how I relate to the ones I love.

Purpose: Auditory training involves active listening to auditory stimuli. Auditory training has been reported to enhance performance in various auditory tasks, supplementing the limitation of hearing aids. The purpose of this case was to determine efficacy of 10-week auditory training for two hearing aids (HA) users even though they have used hearing aids more than 10 years but having little satisfaction on their HAs. Methods: We conducted a 10-week in-laboratory audiotory training (one session per week, about 1 hour per session). Four types of materials (environmental sounds, consonants, sentences, and crossword quiz) were used as training stimuli. Difficulty level for training was adjusted focusing on the use of an adaptive method. To examine the efficacy of the 10-week training, the recognition abilities of environmental sounds, consonants, words, and sentences were compared before and immediately after the training. As a subjective outcome, the Korean Evaluation Scale for Hearing Handicap (KESHH) questionnaire was also conducted before and after training. To examine retention of training effects, we conducted the same evaluations at 10-week, one month and seven months after the completion of training. Results: The recognition scores of all stimuli were improved immediately after 10-week training. The results of the self-reported questionnaire of KESHH demonstrated that auditory training reduced subjective handicap from impaired hearing, indicating that auditory training positively affected individual’s subjective satisfaction of HA. The training efficacy appeared to be maintained even seven months after the completion of training. Conclusion: Auditory training was effective to improve recognition performance as well as the subjective satisfaction of own HA for unsatisfactory HA users.

Purpose: A crossword puzzle is one of the common and familiar activities for adults. The purpose of this study was to develop crossword puzzles as a material of auditory training. In the puzzle, words were answers (target) to the puzzle, and all the target words were placed from left to right and from top to bottom in a crossword grid. As an auditory clue of puzzle, either sentence or a series of words was used to help figure out the answer. Based on pilot study, we selected 112 crossword puzzles, consisting of 503 answer-words (target) words, 503 sentence-hints, and 2,515 word-hints. Methods: To verify validity and difficulty of each item, 20 young and 16 older adults were tested. Using the crossword puzzle, the listener was required to hear word hint and sentence hint that could lead to the answers and say their own answers. Results: Results showed that, in general, the target words were more easily solved with sentence clues compared to series of words. For most of the puzzle items, young and older listeners answered the crossword puzzles correctly. Of 112 puzzles, 12 puzzle items showing less than the score of 80% were removed. Consequently, 100 crossword puzzles were selected as a final item, consisting of 452 answer (target) words, 452 sentence-hints, and 2,260 word-hints. Conclusion: Overall, the crossword puzzles developed in this study could be useful for adults regardless of age. Since the crossword puzzle is familiar and entertaining activity, it might better maintain motivation of auditory training compared to conventional training activity of word or sentence repetition. Further research is needed to generate a larger database of puzzle items and also to determine the feasibility of the crossword puzzles into the clinical purpose of aural rehabilitation for the hearing impaired.

The Simple Temporal Problem (STP) is a fundamental temporal reasoning problem and has recently been extended to the Multiagent Simple Temporal Problem (MaSTP). In this paper we present a novel approach that is based on enforcing arc-consistency (AC) on the input (multiagent) simple temporal network. We show that the AC-based approach is sufficient for solving both the STP and MaSTP and provide efficient algorithms for them. As our AC-based approach does not impose new constraints between agents, it does not violate the privacy of the agents and is superior to the state-of-the-art approach to MaSTP. Empirical evaluations on diverse benchmark datasets also show that our AC-based algorithms for STP and MaSTP are significantly more efficient than existing approaches., Accepted by The Thirty-Second AAAI Conference on Artificial Intelligence (AAAI-18)

난청인은 난청의 정도, 종류, 유형에 따라 각기 다른 정도의 의사소통의 한계를 가질 수 있고 더 나아가 사회적인 고립, 소외감, 열등감 등의 다양한 심리적인 문제까지도 경험할 수 있다. 특히 난청아동의 경우 난청이 단순히 듣지 못함의 문제를 유발할 뿐 아니라 언어, 지각, 인지 능력 등의 저하로 연결되어 더 복합적인 문제가 유발될 수 있다. 따라서 많은 청각장애인들은 원활한 의사소통 및 사회생활을 위해 보청기라는 보조기기의 도움을 받아 소리를 증폭해 들음으로써 도움을 받으며 살아가고 있다. 하지만 고심도 난청인에게는 보청기의 이득이 제한적일 수 있어 최근

© 2016 The Authors and IOS Press. The Simple Temporal Problem (STP) has been widely used in various applications to schedule tasks. For dynamical systems, scheduling needs to be efficient and flexible to handle uncertainty and perturbation. To this end, modern approaches usually encode the temporal information as an STP instance. This representation contains redundant information, which can not only take a significant amount of storage space, but also make scheduling inefficient due to the non-concise representation. In this paper, we investigate the problem of simplifying an STP instance by removing redundant information. We show that such a simplification can result in a unique minimal representation without loss of temporal information, and present an efficient algorithm to achieve this task. Evaluation on a large benchmark dataset of STP exhibits a significant reduction in redundant information for the involved instances.

Objective: In 2012, 4,743 pedestrians were killed in the U.S., representing 14 percent of total traffic fatalities. The number of pedestrians injured was higher at 76,000. Therefore, 36 out of 52 of the largest cities in the U.S. have adopted a citywide target of reducing pedestrian fatalities. The number of cities adopting the reduction goal during 2011 and 2012 increased rapidly with 8 more cities. We examined the scaling relationship of pedestrian fatality counts as a function of the population size of 115 to 161 large U.S cities during the period of 1994 to 2011. We also examined the scaling relationship of non-pedestrian and total traffic fatality counts as function of the population size as well. Methods: For the data source of fatality measures we used Traffic Safety Facts FARS/GES ANNUAL Report published each year from 1994 to 2011 by National Highway Traffic Safety Administration. Using the data source we conducted both annual cross-sectional and panel data bivariate and multivariate regression models. In the construction of the estimated functional relationship between traffic fatality measures and various factors, we used the simple power function for urban scaling used by Bettencourt et al. (2007; 2010) and the refined STIRPAT(STochastic Impacts by Regression on Population, Affluence, and Technology) model used in Dietz and Rosa (1994; 1997) and York et al. (2003). Results: We found the scaling relationship display diseconomies of scale or sublinear for pedestrian fatalities. However, the relationship displays superlinear relationship in case of non-pedestrian fatality. The scaling relationship for total traffic fatality counts display a near linear pattern. When the relationship was examined by the four subgroups of cities with different population sizes, the most pronounced sublinear scaling relationships for all three types of fatality counts was discovered for the subgroup of mega cities with a population of more than 1 million. Conclusions: The scaling patterns of traffic fatalities of subgroups of cities depend on population sizes of the cities in subgroups. Especially 9 mega cities with populations of more than 1 million are significantly different from the remaining cities and should be viewed as a totally separate group. Thus, the analysis on the patterns of traffic fatalities needs to be conducted within the group of mega cities separately from the other cities with smaller population sizes for devising prevention policies to reduce traffic fatalities of both group of mega cities and the other group of smaller cities.

본 연구의 목적은 청소년의 온라인 포인트 기부 참여 실태를 살펴보고, 포인트 기부 참여가 청소년들의 개인적, 사회적 발달에 미치는 효과를 분석하는 것이다. 이를 위해 청소년의 포인트 기부 참여무유와 참여횟수를 살펴보고, 청소년의 개인적, 사회적 발달 변수인 진로성숙, 리더십, 학교적응, 다문화수용도, 공동체의식에 미치는 효과를 경험적으로 분석하였다. 분석대상은 전국 초등학교 4~6학년, 중학교 1~3학년, 고등학교 1~3학년에 재학 중인 청소년들 중층화다단계집략표집에 의해 표집 된 7,052명을 대상으로 하였고, 자기기입식 집단면접조사를 수행하여 자료를 수집하여 분석하였다. 분석결과, 지난 1년간 5.6%의 청소년들이 포인트 기부에 참여한 경험이 있는 것으로 나타났고, 평균 참여횟수는 3.17회인 것으로 나타났다. 그리고 청소년들의 온라인 포인트 기부활동의 참여무유 및 참여횟수에 따라 개인적 발달인 진로성숙, 리더십, 학교적응과 사회적 발달인 다문화수용도, 공동체 의식에 정적인 영향을 미치는 것으로 나타났다. 이러한 결과를 바탕으로 나눔교육 확대, 온라인 기부 콘텐츠의 다양화 등 온라인 포인트 기부 활성화에 관한 방안을 제시하였다.