Cloud Deep Learning
Deep Learning In The Cloud
Miranex design and host scalable Deep Learning platforms in the Cloud, but what is Deep Learning? Part one of a three blog series (links at the bottom of the page for the next installment)
Part 1. What is Deep Learning?
Deep Learning is one of the later buzz words or today’s ICT world and is often seen as part of a wider High Performance Computing arc, yet commonly interchangeably used with AI (Artificial Intelligence) and Machine Learning.
So what is Artificial Intelligence, Machine Learning and Deep Learning in a bit more detail?
Artificial Intelligence (AI), already coined during antiquity and used in mythology to forge the gods, had its renaissance after the emergence of codebreaking and computing machines during the Second World War were there was the creation of the Colossus computer 1943-1945 used for Enigma code-breaking. Thereafter AI research was founded at a workshop during summer 1956 at the Dartmouth College, Hanover – New Hampshire, United States, where the field was also known as fields such as Thinking Machines
Machine Learning and Deep Learning
Machine Learning (ML) is a term that has been around since the 1970s and 1980s, and defines a process where a machine or machines access large data sets with specific pre-programmed algorithms to learn for themselves and become smarter at recognising specific information. They learn from their previous searches how to complete their task quicker or more efficiently, using the algorithm.
Fig. 1 Historic development of computer based learning
Deep Learning is therefore a sub-category of Machine Learning as it takes this functionality a step further. Deep Learning is a new and specific branch of Machine Learning with a focus to solve real world problems and challenges with neuronal networks designed to imitate a specific human decision making process. These trained neuronal networks are then becoming so called Deep Neuronal Networks (DNN). These logic networks are then capable of dealing with very large data sets and identify and classify information with high probability and accuracy at very high speed, yet improve their DNN as they keep learning and sifting through data. Deep Learning however, can be applied to almost any type of information or data, from images, to video to audio, voice and language meaning it has a huge potential in problem solving in the World today and the future.
Some Examples of Deep Learning Applications
Here are a few of many examples of what Deep Learnings can bring:
- Analyse real-time video footage to analyse building, people objects and vehicles. This can help optimise traffic flow, prevent crime and create smarter healthier cities.
- Help prevent global warming poverty by combining satellite imagery with deep learning. By doing this socioeconomic indicators of poverty can be identified and predicted before the traditional reactionary signals are provided.
- Automatic colourisation of black and white photos and films. Using image recognition and very large convolution neural networks, Deep Learning can substitute colours, grading, depth, shadows, and movements
Our Expertise – Cloud Deep Learning, How to Scale.
A Deep Learning test platform, proof of concept (POC) or Minimum Viable Product (MVP) is fairly easy to build and run in a lab, in the office or even at home. However the challenges occur when you try to scale out your Deep Learning platform.
When scaling out the IT hardware costs begin to mount up, along with the electricity demands needed for power and cooling. Additionally, there are highly specialised development tools and surrounding libraries needed to grow and scale-out in a linear fashion in order to cope with increasing demand for storage and performance. This where Miranex can help you develop a tailored solution from a small single workstation for research and development (R&D) purposes all the way to a super cluster with thousands of nodes that scale out to your requirements.
With the assistance of GPU and NVIDIA Telsa and Volta, Miranex is utilising the best in technology and are vendor agnostic which allows us to combine the best components in the IT world or stick to a solution from a single vendor for maximum integration & compatibility of all individual components. The choice and scale is for our customers to decide.
For further information on our Deep Learning products please visit our Deep Learning page, or contact us to discuss your project.
We will continue this series in our next article and cover the question of how to make best use of NVIDIA CUDA and GPU technology in general for Deep Learning, Machine Learning and AI research projects and applications.
2nd Article in this series – “Deep Learning and how to make use of NVIDIA CUDA and GPUs”
3rd Article in this series – “Scale Out and challenges of Deep Learning”
Find out more about our Services.