Build an AI to beat the game Breakout

Main idea to work on this project:

The main idea of me and my Teammate idea was to improvise some project which was already done before.

we got this idea when we were having a coffee conversation about the movie name called BLACK PANTHER , in that a foamous quote says “ if smg works perfectly doesn’t mean it cant be imporoved “

so we both started searching for the topic to do a meaningful project ….. finally we came up with an idea of the Breakout Game. We did some research on that, it was so surprising that it was a very simple game but takes more than ~40 hrs to train the model 🥵 ,we both discussed about this and started searching for solution like whether any algorithm which will reduce the amount of the training time, so that it can be trained with less configuration laptop/desktop

Finally after many research we came up with an algorithm called A3C( Asynchronous Actor-Critic Agents ) which was developed by GOOGLE DEEPMIND what it actually does was this algorithm will use minimum of 3 Actor(Agent), so that the three agent will start their point from different manners and according to their rewards they’ll move on but the main of the 3 actor is they have a common SHARED MODEL through which they choose the optimal part of training the model little bit faster, but this fast is not enough, so we are introducing LSTM layer(Long Short Term Memory layer).

Problem Faced during the implementation:

we have no idea at the starting, but after continously reading many researh papers and videos we came up with an idea like this should be the methdology of implementing this project

But when i take some of the code from Github for creating the environment for the project because we are jsut trying to improve the time complexity of the already done project with and help of the A3C algorithm and a LSTM layer and see if our idea really works

But when i try to run the code on spyder but the mistake which i made was , i started runing the code on the Anaconda environment which i created for Self driving Car project which i created a year back. so that had all the older package of pytorch ,opencv …. which is too old and i started getting error and me and my teammate thinking like what is happening man, we thought it was easy to implement but it is giving error in the starting itself, so we approach our seniors to resolve this error and they found out the error in small period of time and told us to create a new conda environment to work on new project so that you wont have any unknown confussion and gave us some tips of seeing the idea of our project, then we started implementing the A3C algorithm in the Deep Convolutional Neural Network and then add the LSTM layer to the hidden layer….

continued in the next blog😃