The second technical seminar - 14:00 19 May 2019

[chunvv]

The second technical seminar on 19 May 2019

• Content:

  • @sontdhust - Son Tran: Head first convolutional neural network
  • @yuyuvn -- Duc Phan: Bluetooth LE

• Information:

  • Location: Trung's apartment, meguro, tokyo, japan
  • Time: 14:00 19 May 2019

[chunvv]

Rules:

• Keep all devices in silent mode.
• Don't eat/drink on my bed, sitting/lying is ok.
• Don't leave any trash on the floor. (I always have to clean my house after the seminar/workshop)
• On time: starting time, researching time and discussing time
• Share all your knowledge
• Don't disturb presenters/discussion

[chunvv]

Pre-preparation:

• To be updated!

[chunvv]

Materials

Presenter: Son Tran @sontdhust
Topic: Head first convolutional neural network
Slide:

https://metalwhale.github.io/cnn-slide/#slide=1

[chunvv]

Presenter: Duc Phan @yuyuvn
Topic: Bluetooth LE
Slide:

https://docs.google.com/presentation/d/1qnsI-W1BXotahIiTH9A40gyIihHU2m54eU64UUSyzlQ/edit#slide=id.p

[chunvv]

13 members, gather in Meguro (in image + Trung Vu)

Shots:

[dungvt93]

Materials

Presenter: Son Tran @sontdhust
Topic: Head first convolutional neural network
Slide:

https://metalwhale.github.io/cnn-slide/#slide=1

two question for son tran sama
+hyper param batch size effect to speed of trainning but is it effect to accuracy ?
+Don't see dropout layer in slide but dropout layer can be use with convolution layer?

[sontdhust]

Dear @dungvt93, let me answer your two questions:

+hyper param batch size effect to speed of trainning but is it effect to accuracy ?

TLDR: AFAIK, Batch size has no effect to the accuracy, since it is not included in model structure.
Reference

+Don't see dropout layer in slide but dropout layer can be use with convolution layer?

Basically, you can add dropout everywhere you want, it is up to you, even inside the convolutional layers. But you should not do this. Please refer to this post for more information.