Proposal and evaluation of deep exploitation-oriented learning under multiple reward environment

Recently, deep reinforcement learning (DRL) has attracted considerable attention. The well-known deep Q-network (DQN) architecture successfully combines deep learning and Q-learning which is a representative reinforcement learning (RL) method. In general, RL and DRL require many trial-and-error searches. To overcome this limitation, alternative approaches called exploitation-oriented learning (XoL) and deep exploitation-oriented learning (DXoL) have been proposed.Although the effectiveness of DXoL for DQNs has been verified, its effectiveness in an environment where multiple types of rewards are present remains unclear. In this study, we apply the DXoL method to two applications with multiple reward types: the driver drowsiness determination system and the decision-making system. Our experimental results show that DXoL is more suitable for learning priorities among multiple rewards than DQNs in these applications.     

Neural Policy Style Transfer

Style Transfer has been proposed in a number of fields: fine arts, natural language processing, and fixed trajectories. We scale this concept up to control policies within a Deep Reinforcement Learning infrastructure. Each network is trained to maximize the expected reward, which typically encodes the goal of an action, and can be described as the content. The expressive power of deep neural networks enables encoding a secondary task, which can be described as the style. The Neural Policy Style Transfer (NPST)¹ algorithm is proposed to transfer the style of one policy to another, while maintaining the content of the latter. Different policies are defined via Deep Q-Network architectures. These models are trained using demonstrations through Inverse Reinforcement Learning. Two different sets of user demonstrations are performed, one for content and other for style. Different styles are encoded as defined by user demonstrations. The generated policy is the result of feeding a content policy and a style policy to the NPST algorithm. Experiments are performed in a catch-ball game inspired by the Deep Reinforcement Learning classical Atari games; and a real-world painting scenario with a full-sized humanoid robot, based on previous works of the authors. The implementation of three different Q-Network architectures (Shallow, Deep and Deep Recurrent Q-Network) to encode the policies within the NPST framework is proposed and the results obtained in the experiments with each of these architectures compared.     

The unreachable object? Difficulties and paradoxes in the analytical relationship with borderline patients

The author investigates the main difficulties the analyst encounters in borderline patient analysis, focusing on the specific way in which such patients put the analyst's mental functioning to the test and highlighting the most salient elements of the transference-countertransference dynamic. The author picks out several of the paradoxes that characterize the analytical relationship with these patients, who are constantly seeking contact with the object, which is inevitably traumatic for them. On the basis of highly detailed clinical material, the author demonstrates how - no matter which theoretical-clinical model is adopted - a specific technical problem with these patients is how to manage their intense destructiveness. With these patients, countertransferential difficulties are inevitably predominant because of the looming threat of the destruction of the analytical relationship. Maintaining a balance between the recognition-legitimization of primary narcissistic mirroring needs and the recognition-control of narcissistic demands and attacks on the analytical link is as crucial as it is complex. The paper examines the most important therapeutic and anti-therapeutic factors, highlighting the importance of countertransference analysis and self-analysis as ways of accessing as yet unrepresented elements of the patient and analyst respectively. Particular attention is given to the role played by the analyst's subjectivity and to the enactment.     

CONTAINING,TRANSLATING, AND INTERPRETIVE ACTING OUT

In analytic treatment, when patients project unspoken aspects of their internal self and object world, the analyst has to find ways to understand and communicate those expelled phantasies without the patient feeling accused, seduced, or persecuted; even when we do our best at interpreting such inner conflicts, the patient may experience our interpretations as assaults, forcing them to give up themselves or their hope for reconnecting with an object. The patient will resist or fight our efforts through the use of projective identification. Caught up in patient's projections, the analyst in turn may enact some of these phantasies by becoming the object rather than translating its presence in the transference, by overemphasizing one side over another of the patient's conflict, or by interpreting accurately but prematurely. These issues are illustrated in two case presentations and discussed in relation to the views of contemporary Kleinian writers on transference and countertransference.     

Stumbling along in the Countertransference: Following Up Enactments with Balanced Therapeutic Interpretations

In the course of a psychoanalytic treatment, many clinical situations create countertransference pulls or invitations to participate in enactments of various degrees. In these projective identification-based transferences, the patient is often successful in drawing the analyst into archaic object relational patterns of acting out. During these moments, the analyst must struggle to find a way to stay therapeutically balanced. The urge to rush to judgment with punitive, seductive, rejecting, controlling, or manipulative comments rationalized as interpretations must be managed. If these unavoidable countertransference enactments are managed and studied, they can provide useful information about the patient's internal struggles and can show the way to making more helpful and more therapeutic interpretations. Case material is used for illustration.     

Method Acting and Gestalt Therapy with Couples

SUMMARY

Techniques taken from Method acting may be useful therapeutic tools and may lend themselves to adoption by Gestalt therapists. Similarities between the goals and procedures of Method acting and Gestalt therapy are described in some detail. The importance and effectiveness of enacted fantasy are discussed, and techniques of Method acting are described in terms of their relation to Gestalt theory. The importance and application of these techniques to Gestalt therapy with couples are discussed.     

情绪劳动的效应、影响因素及作用机制

情绪劳动是指员工遵照一定的组织规则, 在工作场所与顾客互动过程中进行的情绪调节。研究证明情绪劳动对组织情境中一系列结果变量影响显著, 既有积极影响也有消极影响, 作用对象包括施动者员工、受动者顾客和规则制定者组织。情绪劳动与结果变量的关系受多个员工、工作和顾客特征变量的影响。资源保存理论为情绪劳动的作用机制提供了一种有力的解释。即时、短期与长期效应的整合、指向内部顾客的情绪劳动以及组织氛围的影响是未来研究值得关注的方向。     

Workplace spirituality buffers the effects of emotional labour on employee well-being

This research examines how workplace spirituality buffers the detrimental relationship between emotional labour and subjective well-being among two samples of service workers in the United States and southern China. Drawing on conservation of resources theory, we found that the negative relationship between surface acting and subjective well-being was moderated by workplace spirituality. Specifically, employees with high spirituality were buffered from the harmful effects of surface acting on subjective well-being, whereas employees with low spirituality were not. We found no evidence of a moderating effect on the relationship between deep acting and subjective well-being. These findings shed light on individual differences that influence the emotional labour process and expand our knowledge of cross-cultural similarities and differences in emotion management.     

A visual embedding for the unsupervised extraction of abstract semantics

Vector-space word representations obtained from neural network models have been shown to enable semantic operations based on vector arithmetic. In this paper, we explore the existence of similar information on vector representations of images. For that purpose we define a methodology to obtain large, sparse vector representations of image classes, and generate vectors through the state-of-the-art deep learning architecture GoogLeNet for 20 K images obtained from ImageNet. We first evaluate the resultant vector-space semantics through its correlation with WordNet distances, and find vector distances to be strongly correlated with linguistic semantics. We then explore the location of images within the vector space, finding elements close in WordNet to be clustered together, regardless of significant visual variances (e.g., 118 dog types). More surprisingly, we find that the space unsupervisedly separates complex classes without prior knowledge (e.g., living things). Afterwards, we consider vector arithmetics. Although we are unable to obtain meaningful results on this regard, we discuss the various problem we encountered, and how we consider to solve them. Finally, we discuss the impact of our research for cognitive systems, focusing on the role of the architecture being used.