ICACT20230030 Question.1
Questioner: liuyuya@ucas.ac.cn 2023-02-20 오전 11:58:28 |
ICACT20230030 Answer.1
Answer by Auhor 2101210548@stu.pku.edu.cn 2023-02-20 오전 11:58:28
Chrome  Click!! |
| In the scenario where the pre-trained ANN model is converted to an SNN model, whether the proposed method can still maintain the superiority of performance？ |
Our method is suitable for directly trained pulsed neural networks with LIF neurons, while most ANN to SNN methods first use ANN training weights and then convert to SNN without LIF neurons. Therefore, the ANN to SNN method conflicts with the basic condition (LIF neuron) required by our proposed method, and the two are incompatible. Thanks！ |
ICACT20230030 Question.2
Questioner: mjs9908@g.skku.edu 2023-02-22 오전 11:14:21 |
ICACT20230030 Answer.2
Answer by Auhor 2101210548@stu.pku.edu.cn 2023-02-22 오전 11:14:21
Chrome Click!! |
| hi thank you for your presentation.
i'd like to know what GPU did you use. |
We use NVIDIA GTX1080ti to train Spiking Neural Networks. |
ICACT20230030 Question.3
Questioner: vuvietvu@vnu.edu.vn 2023-02-23 오전 11:48:45 |
ICACT20230030 Answer.3
Answer by Auhor 2101210548@stu.pku.edu.cn 2023-02-23 오전 11:48:45
Chrome Click!! |
| What is the training time for ech experimentation? What is the effect of parameters on the quality of results? |
Thank you for your question. The training time for each experiment was about eight hours. Different parameters have great influence on the experimental results. We have given the parameters with advanced performance and will continue to explore the rule of parameters on the experimental results in the future work. |
