Now 7707 visitors
Today:0 Yesterday:0
Total: 159502 216S 51P 5R
2026-10-22, Week 43
Member Login
TACT Journal Homepage
Call for Paper
Committee
Find My Paper
Author Homepage
Statistics & History
Welcome Message
Paper Procedure
FAQ
Paper Archives
Outstanding Papers
Program & Proceedings
Presentation Platform
Hotel & Travel Info
Photo Gallery
Scheduler Login
Seminar
Archives Login
Sponsors


















Online Q&A Session
Paper Number Author:
Paper Title
Keyword
Q&A Number **Select your question number !!
Questioner eMail:
Question
Answer
by
Author
Save Q&A

* Edit or answer any Q&A by selecting Q&A number Hyper Link below + Write button (Save)
ICACT20260115 Question.1
Questioner: pangb@dsp.ac.cn    2026-02-10 오후 11:35:42 		ICACT20260115 Answer.1
Answer by Auhor aoki.kazuma.s1@elms.hokudai.ac.jp   2026-02-10 오후 11:35:42
Chrome Click!!
When computing the reward, how do you usually choose the value of epsilon? I noticed that in your work, the coefficients for throughput and RTT are both set to 1, and it seems they’re not normalized. In that case, when taking the difference between the new and old utility, the change in throughput could be much larger than the change in delay. After varying epsilon over a wide range, we set it to -10, which yielded the highest average throughput. Throughput always exhibits slight fluctuations, so even during high-throughput periods, Unew – Uold can become negative. To accommodate this, epsilon must be set to a negative value. Furthermore, normalization was not performed this time. The values of beta and gamma were determined based on parameter tuning results under the evaluation conditions used here. This implies that throughput is prioritized over latency, which is correct as you pointed out. Note that during congestion, Unew - Uold < 0. We think that a key future challenge is the need to reproduce stable, high-throughput results independent of network parameters. We are therefore exploring normalizing beta, gamma, and epsilon based on the throughput and RTT terms.
ICACT20260115 Question.2
Questioner: pangb@dsp.ac.cn    2026-02-10 오후 11:36:15 		ICACT20260115 Answer.2
Answer by Auhor aoki.kazuma.s1@elms.hokudai.ac.jp   2026-02-10 오후 11:36:15
Chrome Click!!
In the action design, the parameter alpha is limited to values from 1 to 4. How was this range determined? The alphas range was set as an initial design to validate the proposed frameworks effectiveness. When setting the range, we considered that an excessively large alpha value could cause overly aggressive cwnd increases, potentially triggering frequent congestion. Therefore, balancing stability, we limited it to the range of 1 to 4. Moving forward, we plan to review alphas range and adjustment methods based on the results of this performance evaluation. Furthermore, based on the results of this performance evaluation, we are currently developing a new method to adjust alpha itself.
ICACT20260115 Question.3
Questioner: md.t.a.812@ms.saitama-u.ac.jp    2026-02-10 오후 11:36:33 		ICACT20260115 Answer.3
Answer by Auhor aoki.kazuma.s1@elms.hokudai.ac.jp   2026-02-10 오후 11:36:33
Chrome Click!!
What was the rationale for selecting reinforcement learning over a hybrid or lightweight deep learning model? In this study, we chose reinforcement learning because it enables online machine learning to adapt to dynamically fluctuating network environments. Furthermore, while the DQN used in this study is expected to present computational bottlenecks for practical implementation, since this method targets only sender-side cwnd control, its feasibility is considered relatively high if deployed on the server side rather than edge devices. Additionally, advancements in both hardware and software, such as GPU utilization and lightweight inference processing, are expected to make real-world implementation increasingly feasible in the future.
ICACT20260115 Question.4
Questioner: zhaozx@dsp.ac.cn    2026-02-10 오후 11:36:57 		ICACT20260115 Answer.4
Answer by Auhor aoki.kazuma.s1@elms.hokudai.ac.jp   2026-02-10 오후 11:36:57
Chrome Click!!
1. You mentioned that the Internet is indispensable in modern society and that quality-of-service (QoS) degradations (high latency, low throughput, packet loss) exist. Could you specify what typical application scenarios (e.g., video streaming, cloud computing, IoT communications) are most severely affected by these QoS issues, and why efficient congestion control is critical for them? QoS in the Internet is significantly influenced not by controls at other network layers, but by congestion control at the transport layer. This means congestion control is critical for all applications. Poor congestion control results in reduced throughput, increased packet loss, and higher latency. Reduced throughput negatively affects many applications, such as web browsing and file transfers. As you mentioned, increased packet loss and latency affect diverse applications such as video streaming, VR/AR, and cloud gaming.