Deep Learning
Deep Learning is a subset of machine learning and concerned with algorithms focusing on solving real-world problems by artificial neural networks.
The most recent news that Google’s AlphaGo artificial intelligence (AI) defeated Ke Jie, the world’s number one Go player, revealed a key milestone that AI has reached and demonstrated the ability it can now surpass human experts to find solutions to very difficult real-world problems.
BENEFITS OF DEEP LEARNING
From the unparalleled progress made by deep learning, we can foresee many benefits that can be brought to enterprises by the technology. Here are some examples.
· Fraud detection
E-commencers suffer from frauds in finance and online transaction processing. Scammers continuously scan and exploit system bottleneck, while fraud detection system has to update its algorithms in real-time to defend. The legacy rule-based fraud detection is very difficult to catch up with the fast-changing hack behaviours. However, benefited from the accuracy and fast response of deep learning, enterprises that adapt this technology will be able to deal with the fraud with no human intervention but prompt reaction to stop the fraud, and therefore save enormous cost.
· Cloud Computing System Administration
Nowadays there are thousands of servers in modern data centres and will be more in the future. The explosion of cloud computing has made server management task increasingly important, but also difficult using existing management tools. Deep learning enables system administrators to predict errors and failures and prevent downtime before it happens, rebalance workloads, and save money.
· From Research to Business Recommendation
Deep learning is sparking new insights in a broad range of research fields such as physics, genetics, neural science, medicine. Researchers benefit from deep learning to tackle problems that were once deemed impossible. On-line business giants use deep learning to create personalized recommendations for different buyers based on the history of their browsing and buying behaviours. Deep learning now spreads out to every corner of human life to make impossible possible.
THE ADVENT OF A NEW ERA
Neural networks are not a new concept since invented more than 50 years ago. However, in the last 10 years a couple of important developments have accelerated the utilisation of deep learning in real-world.
The first development is the fast growth of the internet which generated more data that can be used by neural networks for training and producing more accurate models to solve the real-world problems. Today data comes from computers, machines, social media, cloud, and many other sources, making it real Big Data to change everything.
Another development is the advances in algorithms such as deep convolutional neural nets, extending the application of deep learning to areas including image and sound recognition, machine translation, and predictive analytics.
The last but not the least, the boost of hardware especially GPU capabilities brings the computing power to a much higher level which is necessary to train deep neural networks.
Combining the three key factors, the deep learning has reached an unprecedented era to catalyse businesses by its power.