Reflections on practical deep learning lesson 1

An Intro to practical deep learning

I am currently taking the fast ai course on Practical deep learning , and I have been impressed by the knowledge I am gaining. Given the explosion of AI, GPTs, and Deep reinforcement learning, I have found that understanding the main driver of it all (Neural Networks) to be quite important for it.

Simply put, Neural Networks are computerized ways for to develop solutions similar to how our brain works. The brain is a bunch of neurons connected together that get ignited as information as being relayed.

With computer neural networks, we utilize an artificial neuron, and create multiple layers for them to work on a task together. This is obviously under machine learning but the real power is that those neurons break the task into their simplest form and try to find a solution to it while cooperating together.

Neural networks is where the idea of deep learning stems from because the deeper the layers provided the more the neurons can combine to solve a problem. I will add that there is a concept of creating solutions similar to the actual entity the neural network is predicting.

Application of neural networks:

Neural networks are applied in multiple fields and industry:

  • Medicine: Using image classification have been helpful in identifying malign tumor and cancers in the medical field. There are many examples that can be found online.

  • Automated driving: Using neural networks (particularly with computer vision) have allowed engineers create vehicles that can “see” and navigate by itself. Obviously, this might need some human supervision but have been found to drive themselves

  • Medicine discovery: NN have been used to discover new patterns with drug molecules that have led to some significant breakthrough.

  • Natural Language Processing: NLP which is the understanding of text information, and generation of new text utilizes neural network to understand the probability of words together. LLMs e.g ChatGPT, SMS completion, Text correction (e.g grammarly) all utilize neural networks one way or the other.

  • Recommendation systems: Recommendation engines such as netflix movie recommender, Amazon shopping recommender, Tiktok video recommender all use neural network in one way or the other to ensure what the user wants they can get based on their previous interactions.

More of the class can be found here.