Repetition-Based Approach for Task Adaptation in Imitation Learning

Transfer learning is an effective approach for adapting an autonomous agent to a new target task by transferring knowledge learned from the previously learned source task. The major problem with traditional transfer learning is that it only focuses on optimizing learning performance on the target task. Thus, the performance on the target task may be improved in exchange for the deterioration of the source task's performance, resulting in an agent that is not able to revisit the earlier task. Therefore, transfer learning methods are still far from being comparable with the learning capability of humans, as humans can perform well on both source and new target tasks. In order to address this limitation, a task adaptation method for imitation learning is proposed in this paper. Being inspired by the idea of repetition learning in neuroscience, the proposed adaptation method enables the agent to repeatedly review the learned knowledge of the source task, while learning the new knowledge of the target task. This ensures that the learning performance on the target task is high, while the deterioration of the learning performance on the source task is small. A comprehensive evaluation over several simulated tasks with varying difficulty levels shows that the proposed method can provide high and consistent performance on both source and target tasks, outperforming existing transfer learning methods.

Clarifying Optimum Setting Temperatures and Airflow Positions for Personal Air Conditioning System on Flight

In recent years, the demand for air travel has increased and many people have traveled by plane. Most passengers, however, feel stressed due to the limited cabin space. In order to make these passengers more comfortable, a personal air-conditioning system for the entire chair is needed. This is because the human body experiences discomfort from localized heating or cooling, and thus, it is necessary to provide appropriate airflow to each part of the body. In this paper, a personal air-conditioning system, which consists of six vertically installed air-conditioning vents, will be proposed. To clarify the setting temperature of each vent, the airflow around the passenger and the operative temperature of each part of the body is investigated using fluid simulation. In the simulation, the ideal temperature for each part of the body is defined and compared with the operative temperature to verify how close both temperatures are, resulting in determining the ideal setting temperature. The simulation result shows, that most parts of the body reach their ideal temperatures. In addition, the optimum setting temperature and position of each air-conditioning vent, which contribute to maintaining the thermal comfort of the human body on the plane, is clarified.

Simulation-Based Designing of Suitable Stimulation Factors for Presenting Two Phosphenes Simultaneously to Lower Side of Field of View

Using a phosphene has been discussed as a means of informing the visually impaired of the position of an obstacle. Obstacles underfoot have a risk, so it is necessary to inform the visually impaired. A previous study clarified a method of presenting phosphene in three directions in the lower vision; however, the simultaneous presentation of these phosphenes has not been discussed. Another study discussing the effect of electrical interference when stimulating the eyeball with multiple electrodes indicated that it is important to select appropriate stimulation factors to avoid this effect. However, when the stimulation electrodes are arranged remarkably close, there is a high possibility that the stimulus factor presented in the previous study will not apply. In this study, a method for simultaneously presenting phosphenes in the lower vision is presented. The electrode arrangements reported in the previous study to present phosphene in the lower field of vision are used, and the difficulty in the simultaneous presentation of multiple phosphenes in the lower vision is the focus. In this paper, the method of designing the stimulation factors is discussed numerically when the electrodes are arranged remarkably close. As a result, it is shown that stimulation factors different from the previous research were appropriate depending on the distance between the electrodes.

Domain Adaptation for Imitation Learning Using Generative Adversarial Network

Imitation learning is an effective approach for an autonomous agent to learn control policies when an explicit reward function is unavailable, using demonstrations provided from an expert. However, standard imitation learning methods assume that the agents and the demonstrations provided by the expert are in the same domain configuration. Such an assumption has made the learned policies difficult to apply in another distinct domain. The problem is formalized as domain adaptive imitation learning, which is the process of learning how to perform a task optimally in a learner domain, given demonstrations of the task in a distinct expert domain. We address the problem by proposing a model based on Generative Adversarial Network. The model aims to learn both domain-shared and domain-specific features and utilizes it to find an optimal policy across domains. The experimental results show the effectiveness of our model in a number of tasks ranging from low to complex high-dimensional.

Generative Adversarial Network for Imitation Learning from Single Demonstration

Imitation learning is an effective method for training an autonomous agent to accomplish a task by imitating expert behaviors in their demonstrations. However, traditional imitation learning methods require a large number of expert demonstrations in order to learn a complex behavior. Such a disadvantage has limited the potential of imitation learning in complex tasks where the expert demonstrations are not sufficient. In order to address the problem, we propose a Generative Adversarial Network-based model which is designed to learn optimal policies using only a single demonstration. The proposed model is evaluated on two simulated tasks in comparison with other methods. The results show that our proposed model is capable of completing considered tasks despite the limitation in the number of expert demonstrations, which clearly indicate the potential of our model.

An Experimental Study of the Server-based Unfairness Solutions for the Cross-Protocol Scenario of Adaptive Streaming over HTTP/3 and HTTP/2

Since the introduction of the HTTP/3, research has focused on evaluating its influences on the existing adaptive streaming over HTTP (HAS). Among these research, due to irrelevant transport protocols, the cross-protocol unfairness between the HAS over HTTP/3 (HAS/3) and HAS over HTTP/2 (HAS/2) has caught considerable attention. It has been found that the HAS/3 clients tend to request higher bitrates than the HAS/2 clients because the transport QUIC obtains higher bandwidth for its HAS/3 clients than the TCP for its HAS/2 clients. As the problem originates from the transport layer, it is likely that the server-based unfairness solutions can help the clients overcome such a problem. Therefore, in this paper, an experimental study of the server-based unfairness solutions for the cross-protocol scenario of the HAS/3 and HAS/2 is conducted. The results show that, while the bitrate guidance solution fails to help the clients achieve fairness, the bandwidth allocation solution provides superior performance.

P-603 Presence of an asymmetrical response to ovarian stimulation in patients with low ovarian reserve

Study question

Does ovarian reserve decline with a symmetrical manner between right and left ovaries in poor responders (POR) with diminished ovarian reserve (DOR)?

Summary answer

Asymmetrical ovarian response to ovarian stimulation with the left-side dominance was found in POR with DOR.

What is known already

Ovarian follicles are produced during fetal stage and not regenerated after birth. Thus, the number of ovarian follicles declines with age, resulting in infertile POR with DOR. In the morphometric study of human neonatal ovaries, no significant difference was found in the number of follicles between the right and left ovaries in the same individual. A previous study demonstrated that there is a difference in the number of follicles between right and left ovaries in patients with normal ovarian reserve with the right-side dominance, suggesting the asymmetrical activation and growth of follicles.

Study design, size, duration

A retrospective analysis was conducted in patients with POR with DOR based on the Bologna Criteria. Inclusion criteria was patients who received more than five times of ovarian stimulations followed by oocyte retrievals. Data were obtained from a total of 265 participants who received IVF-ET treatments from April 2015 to March 2021 after receiving written informed consents under an approval from the institutional ethical committee. Patients with the history of previous ovarian surgery were excluded.

Participants/materials, setting, methods

The enrolled patients were received ovarian simulation under short or GnRH antagonist protocols for oocyte retrieval. We collected the data of retrieved oocyte number as well as the outcome of IVF from medical chart. We defined the right-left asymmetry of ovarian reserve (%) based on the number of retrieved oocytes from dominant side ovary per total number of retrieved oocytes. Statistical significance was determined using Dunnett or chi-square tests, with P < 0.05 being statistically significant.

Main results and the role of chance

The average age of participants was 37.2±5.99 years of age exhibiting low serum AMH levels (average 0.09±0.20 ng/ml). We analyzed 2,181 cycles of ovarian stimulation (average 8.3±3.9 cycles/patient). The number of retrieved oocytes were 3, 882 in total cycles (average 12.8±7.1/patient). Among participants, 22 cases (8.4%) showed left and right equal in the number of retrieved oocytes, whereas >70% asymmetry was observed in 107 cases (40.7%) and >80% asymmetry was detected in 60 cases (22.8%). In 18 cases (6.9%), oocytes were collected from one side ovary only showing 100% asymmetry. In the cases with >70 and 100% asymmetry, the left-side dominance was 1.3-fold and 5.0-fold higher than right-side dominance, respectively. In cases with 100% asymmetry, there was no difference in the number of cryopreserved high-quality embryos between left and right sides of ovary.

Limitations, reasons for caution

Although we enrolled POR with DOR patients who received ovarian stimulations more than five times, the duration of ovarian stimulation was different among patients. It affects the numbers of ovarian stimulation cycles and retrieved oocytes in each patient.

Wider implications of the findings

Considering the finding of right-side dominance in the number of follicles with normal ovarian reserve, the activation and development of follicles might be accelerated in the right side due to asymmetric blood supply to the ovaries, and thus follicles are likely remained in the left-side ovary with low ovarian reserve.

Trial registration number

not applicable

SVD-Based Mind-Wandering Prediction from Facial Videos in Online Learning

This paper presents a novel approach to mind-wandering prediction in the context of webcam-based online learning. We implemented a Singular Value Decomposition (SVD)-based 1D temporal eye-signal extraction method, which relies solely on eye landmark detection and eliminates the need for gaze tracking or specialized hardware, then extract suitable features from the signals to train the prediction model. Our thorough experimental framework facilitates the evaluation of our approach alongside baseline models, particularly in the analysis of temporal eye signals and the prediction of attentional states. Notably, our SVD-based signal captures both subtle and major eye movements, including changes in the eye boundary and pupil, surpassing the limited capabilities of eye aspect ratio (EAR)-based signals. Our proposed model exhibits a 2% improvement in the overall Area Under the Receiver Operating Characteristics curve (AUROC) metric and 7% in the F1-score metric for 'not-focus' prediction, compared to the combination of EAR-based and computationally intensive gaze-based models used in the baseline study These contributions have potential implications for enhancing the field of attentional state prediction in online learning, offering a practical and effective solution to benefit educational experiences.

Evaluation of Phosphene Shifts During Eye Movements to Enhance Safe Visual Assistance for Visually Impaired Individuals

Hands-free visual assistive devices that consider the safety of the visually impaired have been researched, but many of them interfere with other senses, such as hearing. Therefore, phosphenes have been researched as a method of presenting visual information. Phosphenes are flashes that are recognized by electrical stimulation, and the presentation position can be adjusted by the electrode arrangement. However, it has been reported that the presentation position changes significantly when the eyeballs are moved as far left and right as possible. As a walking assistive device for the visually impaired, the fluctuation of the presentation position of phosphenes may cause safety problems. This study used the Hess test to verify the positional fluctuation of phosphenes associated with eye movement and compared it with eye movement during walking to discuss safety. As a result, the range of the change in the presentation position of phosphenes was significantly large when subjects moved their eyes to the peripheral vision, the same as in the previous study. On the other hand, the presentation position of phosphenes did not change significantly within the range of eye movement during walking (±15 deg). Our results suggest that there is the possibility that serious safety issues will not happen with assistive devices for the visually impaired who use phosphenes.