Neural networks are based on the philosophy from the brain; however, the brain consists of around 100 billion neurons with each neuron connected to 10,000 other neurons. Neural networks developed in the early 90s faced a lot of challenges in building deeper neural networks due to computation and algorithmic limitations.

With advances in big data, computational resources (such as GPUs), and better algorithms, the concept of deep learning has emerged and allows us to capture a deeper representation from all kinds of data such as text, image, audio, and so on.

• Trends in Deep Learning: Deep learning is an advance on neural networks, which are very much driven by technology enhancement. The main factors that have impacted the development of deep learning as a dominant area in artificial intelligence are as follows:

• Computational power: The consistency of Moore's law, which states that the acceleration power of hardware will double every two years, helped in training more layers and bigger data within time limitations
• Storage and better compression algorithms: The ability to store big models due to cheaper storage and better compression algorithms have pushed this area with practitioners focusing on capturing real-time data feeds in the form of image, text, audio, and video formats
• Scalability: The ability to scale from a simple computer to a farm or using GPU devices has given a great boost to training deep learning models
• Deep learning architectures: With new architectures such as Convolution Neural network, re-enforcement learning has provided a boost to what we can learn and also helped expedite learning rates
• Cross-platform programming: The ability to program and build models in a cross-platform architecture significantly helped increase the user base and in drastic development in the domain
• Transfer learning: This allows reusing pretrained models and further helps in significantly reducing training times