₦airaland Forum

Marvel1206:
So how did you start learning this?

Learning Python was the first step. Microsoft did an online lecture about python around that time and I followed on Youtube. Once I was comfortable with Python I got the book Hands on machine learning with Scikit learn and Tensorflow which gave me good introduction to machine learning with each chapter about a different kind of ML algorithm/Model (I read many other books but none is as influential as this book). Subsequently i took a few courses on Udemy about Reinforcement learning that introduce me to the field.

Kyase:
On the contrary, Tinubu handled it the way it should be, enough of trying to be the nice guy.

Why should he prove himself to anyone, Tinubu is not the president and at such can't order the army to shoot ( I'm doubting the lekki thing sef)

What I saw in Tinubu yesterday is a fearless Man, stamping his ground.

Tinubu just won himself a fantastical fan and supporter in myself.

Ride on jagaban

You mean he just won himself a dark souled miscreant...

Only God knows how this people sleep at night. Power indeed intoxicates.

gidson12:
this ia CACOVID palliative, where is federal Government palliative,

She doesn't know the difference bro.

Predstan:
Interesting. I am also an AI enthusiast and my interest is on the Microcontroller programming (TinyML). Interestingly I don’t like Pytorch, I use Tensorflow for deep leaning and neural network.

I started with Tensorflow actually, but it doesn't feel pythonic to me so I switched to pytorch but with the Keras api now merged with the new Tensorflow, I think I might start to use it again.

I’m currently working a traffic light(Computer Vision) model to be deployed to PyBoard and Arduino.


Sounds interesting, could you explain more about it. And why PyBoard and not raspberry pi.

A country with good IT infrastructures. Not like the ones anonymous is toying with.

helinues:
No one is disputing the soldiers shooting on protesters ...

The massacre is what we are disputing

And this is how psychopaths are made. Check your soul bro.

Conclusion

The experiment was a bit of success although the agent didn’t maximize its potential mainly because it can only sense the board only through the region around it’s head and a better option is the snake seeing the whole board as an image through a convolutional neural network. The agent was able to eat about 50 apples at max.

For the next experiment, I will be training an agent to play the old shooting game CONTRA and I am really excited about the training. I will also build upon the lesson learnt by this experiment by giving the agent more awareness of the environment by providing the agent the direct image render of the screen. IF the next experiment is successful, other prospective games for experimenting include Asphalt, Mortal Kombat etc. and also maybe soccer (but definitely the old versions like winning eleven 2008 and will probably need the help of a C++ dev to tune the game to a usable environment)
Adios guys.

The loss function is the hubber loss. A hybrid of the Mean square error loss and Mean absolute error loss function. Another viable loss function is the trusted Mean square error loss itself.

The agent itself is nothing but three layers of feed forward neural networks with 64 hidden layers, nonlinear activation functions, and an output of three choices LEFT, RIGHT or keep moving STRAIGHT.

For the training loop, the environment provides the agent with a state containing vital information the agent needs to navigate the board. The state includes the direction of the location of the apple (i.e. north, west, south, east), the current direction the snake is travelling, and details of the surrounding position of the snake’s head whether they are occupied or not.

The reward function for the training is the same as the one described above, but I added an extra negative -0.05 reward for each action the agent takes that doesn’t result to the snake eating the apple. This will tell the agent to eat the apple as soon as possible and not to roam around the board.

The agent is trained using the principles of Q Learning, Bellman Equation and the Markov’s decision process. The above is used to exploit an environment with a number of states by an agent performing an action on the environment. In this experiment, the environment is the snake game.

The environment gives our agent a reward for each action performed. The reward is positive when the agent eats the apple, and negative when the agent crashes the snake on the wall or when the agent crashes the snake on itself. The purpose of this reward is to learn a good policy i.e. eating apples is good while crashing is bad.

And here is the agent in it’s infancy stage about 0 – 2 hours of training. It can be seen performing random actions.

[img]https:///media/10-24-2020/--Uk7D.gif[/img]

The training took about 26 hours on an hyper threaded Intel Xeon CPU (good reward aware policy was learned around 5 hours). Although I tried training with a free tesla K80 GPU from google but I was getting almost the same result from my benchmark because most of the work the agent is doing is on the CPU.

In this post, I will be showing the result of my experiment with an AI agent trained with Pytorch that plays the game snake. Here is the final result.

[img]https:///media/10-24-2020/Jax6Kp.gif[/img]

During this past few weeks I have been able to explore another part of the AI promise, automated gaming.
Games are part entertainment and part logical and strategical so training an AI agent through reinforcement learning could help us develop other important agents like the case of an agent trained by google that was able to cut their electricity bill of a server building by more than 70%.

Self-driving vehicles, Industrial scale automations, Operations in unsafe zones like oil rigs, Increased successes in surgical operations … these are part of the many promises of artificial intelligence.

Do you know that the highest grandmaster of the hardest board game in the world GO is an AI? 9 March 2016 – 15 March 2016 was the week AlphaGo broke the internet by beating Lee Sedol an 18-time world champion by 4 – 1. It was a feat considered impossible as the number of possible moves in the game of GO is more than the number of atoms in the universe. And that was the week I was seduced from the medical career I was pursuing and everything became history.

TLDR: This writeup is for enthusiasts of AI, automated gaming, and application’s of machine (deep) learning in general.

Fuckingmallam45:
When you’re all ready to learn, you people should let me know . This time around, you should all talk with courtesy and a lot of respect. Because we are all on a Faceless forum, una Dey always think say we be mate.

In all sincerity, I really want to learn. So could you please explain how it is possible?

clintz:
learn to understand people, videos of dead protesters being paraded on social media are fake

Fuckingmallam45:
Photoshop is an Adobe product to manipulate and edit photos. There’s an another Adobe product named Adobe After Effects that can be used to edit and manipulate LIVE videos.
Say what you know.

Both of you are royally stupid. Saying rubbish with confidence. So you are saying they somehow infiltrated Instagrams server and they manage to doctor the videos before the video return trip to the many clients streaming it?

If they can manage to doctor such videos why is it that the vfx in our mainstream movies are still whack?

As i always say, CNN is a stupid media. I'd prefer AIT over them any day? Clash between protester or massacre of protesters?

Another one term governor!! We won't forget come 2023, you are out!!!!! We won't forget this.

crackhaus:
Well I guess you've been following this jailbreak news since morning, did you hear anything about casualties or a gun duel?

Did you even hear any gunshots in any of those videos?

I think they realize people are starting to figure out it's an inside job, and all these stories are coming out now more than ten hours after the incident happened.
Makes no damn sense...

Prisoner wearing chain One was carrying travelling bag, one with dr dre headset. Is the chain not even a potential attack weapon and why is he allowed to wear it. Scam prison break...

How about oyinbos that think africa is a country? Shitty media's like cnn do it regularly!!

AntiBalaka:
Hack Buhari's twitter.

Annonynous are Glow Niggers by the way.

They are

C
I
A

One hot akamu for this guy. They are either the agency or the N o S uch A gency.

Took me 9hrs from the island to my house last week. Spent about 5 hours alone on the 3rd m bridge. Its frustrating yes, but all is fair in love and war.

Beverlyjean:
this funny isaac Newton story that has been told for decades...was apple the first time he ever saw something that drop ...

Eureka moment. It happens.

Chrisjay1998:
So where is the Prove hes a Corrupt dirty individual
even if he's Corrupt at least he came out to Voice out on the Sufferings of Nigerians
Your mother must have been a foolish person to have given birth to you
You Should have been Aborted Period !!!!

When you don't know something at least search, don't come on the internet to display your low intelligence. I will help you a bit, go read about Noah Glass.
For your other ramblings, you are probably suffering from traumatic brain injury so I only pity you. Go get help lowlife.

The pawn. They are very underated, meaning stealth attacking my opponent with the pawns without him knowing whatsup until too late. Also, its the only piece that can evolve into any other piece except the king.

Jack too is a corrupt dirty individual. He shouldn't be claiming one idiot activist.
#ENDCORRUPTION