sentdex
Setting up and visualizing our data for the convolutional neural network.
#machinelearning
#deeplearning
#python
Source
Structuring and visualizing Data – Deep Learning in Halite AI competition p.2
Similar Posts
Miguel Encarnao – Intelligent Caring Machines: FHTI Virtual Healthcare Workers Initiative fhti.org
What is Consciousness? – J.P. Moreland
The Artificial Intelligence Breakthroughs of The Last Five Years – Rob Fergus
26 thoughts on “Structuring and visualizing Data – Deep Learning in Halite AI competition p.2”
Comments are closed.
First!
That is awesome! Thanks for the content!
So early!
I wonder, who the hell gave dislike on this video! Some people are really weird, they don't know how to respect other's knowledge
Hello
Please sir can you send me the playlist of the series you making now and the one before this series I got lost of the following after two weeks off the tutor… I stopped at python 3 lecture 6
It kinda amazes me that you do all of these stuff through simple and rudimentary code. Your code has always been easy to understand.
10:34 – Yes you do indeed deposit it. This can be used during the endgame as a quick way to deposit up to 4 ship's halite in one turn – one ship comes from each cardinal direction and collides on the shipyard.
You will deposit the halite if you die on your dropoff. It's actually a very common strategy to "suicide" your ships onto the dropoff before the game ends.
Hi Harrison, I hope you are well mate. would you mind answering some questions for me about my job hunting state?
this is my situation:
I am looking for a job for two years as a data analyst or data scientist and I have not got anything. My academic field is not computer sciences but economics. During the last two years, I am developing my python and SQL skills. I have learned how to implement machine learning models and clean data with python. Although I feel that I still have a long way to do.
I am from Spain so I moved to London because I think that it would be easier to put my foot in some company over there. I did at least 15 interviews, some of them were second interviews for the same role. I was interviewed in some big financial companies as well as in small companies. Some interviews went really good and I got good feedbacks but eventually, I got rejected.
On the other hand, when I am doing some analytics at home with my laptop, I find a lot of obstacles over the way such as errors in my software during the code (pycharm) or a lot of steps that i don't know what they are for. Which make me stop doing what I am doing because I have no motivation as I have no job.
I would be so pleased if you could give some advice about this. I am sure that for a guru in data scientist like you it is very easy.
many thanks mate!!
Halite on a map cell is not capped, so after two ships collide the halite amount dropped can exceed 1000.
The file is in append mode
Do you let your stuff still run on an Intel CPU or could AMD convice you?
How do you plan to get any direction/localisation from inputting the map to a conv net? I am still a newbie, but my understanding of a convolution is that the features extracted are invariant to position in image? What am I missing?
More descriptive variable names could help. I don't know how you remember what your code does. Great video
23:30 lol man, I see my self in those situations, such a fun to watch from the side hahahah
There was an ? minute ad before I watched this, what the f#$&
Edit: I went to another video and when I watched this video again the ? minute ad appeared again. Wtf, and the ad is the full LEGO movie, even more wtf.
This is really interesting, I'm very much looking forwards to the next one in this series, keep up the good work my man!
what we have done in 34:42 exactly? what is 'secrets' there?
I came across this bot which is designed to park itself in the enemy drop off mentioned at 10:07 . https://www.youtube.com/watch?v=Cu6iJvctnhs&lc=z23yyfcwetrks5s5j04t1aokgk2punas4otpms1hytxmrk0h00410.1542604467652097
Great stuff. One idea, I guess the way we display ships is not optimal because we can only recognize them if they have already collected halite otherwise the look the same as an empty cell. Maybe we could try to normalize the ships between 0.5 and 1 instead of 0 to 1?
Is it me or does the profile picture get more and more pixelated?
10:30 Hey, Don’t know if you saw this replay (https://halite.io/play/?game_id=687791&replay_class=1&replay_name=replay-20181018-073753%2B0000-1539848266-40-40) on halite’s forums, but yeah at least I along with several people hadn’t thought of enemy ships being on our own shipyards.
Harrison, just do game_map.width to calculate the width and then no matter the map size you can still see the whole map
For me its saying that the position in {ship.position + Position(…)} is not defined. Someone knows whats going on ?? or someone is having the same problem ???
i feel like there should be a faster metho for building the surroundings numpy array – especially later if we wan to use a model with >50 Ships per turn. Maybe somebody has an idea using numpy?
which software / idle is this ?