The first question someone would ask me, is why AI ?
I have always loved the Iron Man movies, and I’m sure most of you do too. JARVIS, who is iron man’s AI assistant, was the first thing that amazed me, since it was the first time I had come across something like this. Then, Apple, created SIRI, Microsoft created CORTANA, and so on.
I started playing around with these AI models, and researched more about them during my free time. That was the time I realised, I wanted to take my interest in this subject to the next level. That is when I started learning Python, one of the main programming languages used in Artificial Intelligence and Machine Learning.
During my summer holidays, in July, 2017, I got an internship at http://www.hotify.ai/ , a company that follows the stream of AI. I was introduced to the CEO and, at my first session, he told me about AI, and gave a thorough understanding of how it works.
I was introduced to a website called Open AI gym, where one can train different environments/games to learn from their previous failures. I was so excited to solve a problem like this.
Before starting a problem, researching about it is always important. I was introduced to terms like “hyperparameters”, “reinforced learning”, “neural networks”, which I had no idea about. When I actually went online and read about it, it all made sense to me, and then I knew how to proceed.
During my first week, I created the basic model, and got it working using a basic neural network. My second task, was to use a hyperparameter and make the model learn through reinforced learning. A hyperparameter consists of a) The learning rate of your model, b) Two layers, an input layer and output layer, c) A function you need to create on python, called activation, d) A loss function.
LR = 1e-3 This is the learning rate
Input Layer (Layer 1)
network = input_data(shape=[None, input_size, 1], name='input') #rectified linear unit (relu) #All 5 can satisfy a 128, but I just tested it ot see what happens, there's no change network = fully_connected(network, 128, activation='relu') network = dropout(network, 0.8) -> This is an example of my input layer
Output Layer (Layer 2)
network = fully_connected(network, 2, activation='softmax') network = regression(network, optimizer='Adam', learning_rate=LR, loss='categorical_crossentropy', name='targets') #Adam is a type of optimizer model = tflearn.DNN(network, tensorboard_dir='log', ).
It is amazing how you can train a computer to play a game, in just a few days. Open AI gym, is a great place for young learners to explore Machine Learning, and I am glad I got introduced to it. My dream is to make an AI model of my own someday, and I thank Hotify for giving me the exposure to AI.
When I was working, I used to be so indulged in this work, I would lose track of time sometimes.
This was my first real experience with Artificial Intelligence and Machine Learning, and it was amazing! I hope all young AI lovers, can explore more about this field, and make it either your hobby, or career interests, like I did.
At TowardsAI, we are especially interested in discovering young talent, and support them (through our network) to prosper in their chosen fields of AI research. Therefore, we heartily request you, your kids, school and college students, to use the 'Comments' section below, to say Hi! to Aditya, and thereby help us build a community of young learners in AI. Thank you for your enthusiasm. :-)