2017 agents: papers, codes, comments

[Dejan Bozhinovski]

Hello everyone,

I was not able to find any info on the agents that were top places for both tracks for 2017 contest.

I could only find the papers and codes from 2016 winners.

I searched through everything and still can't seem to find them.

I am interested to find their implementations, a paper, or in case of not published paper, because lack of novelty, at least a short comment on what were the techniques used.

I am really curious what is behind the brain of the bots Marvin, and Yanshi from TSAIL.

Best regards,

Dejan

[Devendra Chaplot]

Hi,

We won the Full Deathmatch track in the 2017 contest. 

The code is available at https://github.com/glample/Arnold.

The paper is at https://arxiv.org/abs/1609.05521.

The paper contains the architecture of the model. There were some changes after the paper, such as increasing the action space, which you can see in the code.

Devendra

--
You received this message because you are subscribed to the Google Groups "Visual Doom AI Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to vizdoom+unsubscribe@googlegroups.com.
To post to this group, send email to viz...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/vizdoom/429f0d7a-4d9f-4c45-85f3-9e487cc4ec61%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

Devendra Chaplot

[Dejan Bozhinovski]

Hi Devendra,

thank you very much for the reply. 

Your paper and code is a very beneficial source for studying.

Are you going to participate this year as well?

Is it possible to find information about the other teams agents?

On Saturday, April 7, 2018 at 8:49:58 PM UTC+2, Devendra Chaplot wrote:

Hi,

We won the Full Deathmatch track in the 2017 contest. 

The code is available at https://github.com/glample/Arnold.

The paper is at https://arxiv.org/abs/1609.05521.

The paper contains the architecture of the model. There were some changes after the paper, such as increasing the action space, which you can see in the code.

Devendra

On Fri, Apr 6, 2018 at 6:51 PM, Dejan Bozhinovski <dejan...@gmail.com> wrote:

Hello everyone,

I was not able to find any info on the agents that were top places for both tracks for 2017 contest.

I could only find the papers and codes from 2016 winners.

I searched through everything and still can't seem to find them.

I am interested to find their implementations, a paper, or in case of not published paper, because lack of novelty, at least a short comment on what were the techniques used.

I am really curious what is behind the brain of the bots Marvin, and Yanshi from TSAIL.

Best regards,

Dejan

--
You received this message because you are subscribed to the Google Groups "Visual Doom AI Users" group.

To unsubscribe from this group and stop receiving emails from it, send an email to vizdoom+u...@googlegroups.com.

To post to this group, send email to viz...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/vizdoom/429f0d7a-4d9f-4c45-85f3-9e487cc4ec61%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

Devendra Chaplot