Mar 1, 2016

The Role of Deep Learning in Data Science

Deep learning: have you heard of the term?

Those with some interest in computer science are most likely familiar with the term, and have read a few lines about neural networks. Otherwise, you are probably asking: what is deep learning? Why should I be interested? And how is this new form of Artificial Intelligence (AI) affecting my life?

What is deep learning?

The term belongs to the family of machine learning methods, which belongs to the field of artificial intelligence. It uses a model of computing called a neural network which mimics the human brain. Human brains work thanks to neurons, which talk to each other. By feeding the computer with algorithms and data, scientists turn computers into intelligent brains to improve natural language processing, speech recognition, image recognition, and more.

Allow us to take a simplistic step back for a moment. To better understand the meaning of deep learning, you must separate it from the origins of machine learning. This earlier discipline often had to do with supervised data. Programmers fed the computer with data and the complex rules to classify the data, which was then used to recognize speech or objects.

Let’s fast forward now. Deep learning uses unsupervised data. There’s no need to feed the computer with how to classify or treat the data, because it learns and thinks on its own based on the algorithms of the deep learning scientist. It starts gaining experience by using multi-layers in complex structures, just like our brain neurons do.

Deep learning techniques require high performance computing power to process the massive amounts of unstructured data through deep learning frameworks and algorithms. Up until recently, high performance computing (HPC) was only available to the largest budgets for hardware and talent to optimize scientific codes on HPC platforms.

Since 2006, the graphic processing unit (GPU) hardware has emerged from more expensive compute intensive graphics applications to more affordable general compute devices available to any compute intensive applications such as deep learning.

Which companies use deep learning technologies?

Many you’ve likely heard of — and others emerging recently — including: Apple, Baidu, Enlitic, Facebook, Google, IBM, MetaMind, Microsoft, Nervana Systems, PayPal, Pinterest, Skymind, Twitter, United Technologies, Yahoo, and many more.

What are the usual applications?

— Natural language processing — This is the process of having a computer understand what you are saying; useful for translation, discourse analysis and text summary.

— Image recognition — This allows images to be easily searched, sorted, and segmented for object detection; which is highly useful to scientific applications, e-commerce, realtors, social networking and advertising.

— Speech recognition — This is when you tell your car to select a radio station, tell your smartphone to type text, or tell your phone to dial a phone number.

So, are neural networks only for large companies?

Certainly not! From automotive, aerospace and finance industries to filmmaking and medical diagnostics, deep learning technologies are applied to all fields — in companies of all sizes.Enlitic, a healthcare company, uses this technology to improve diagnostics. Medical data is screened to provide the patient with detailed data relating to potential diseases.

Cellscope, a pre-IPO company, developed an iPhone otoscope which helps parents to diagnose the condition of their children suffering with acute otitis media instead of repeatedly visiting the doctor.

Are you looking for the right deep learning scientist?

The key is to focus on the application area you’re most interested in. Do you want speech or object recognition? Look for a candidate specializing in this field. The overall requirements may also include:

— A PhD in computer science and specific data science specialties. Training, experience, ability to handle and analyze big data, and develop algorithms with proficiency in certain computer programming technologies.

— A PhD that has graduated from universities with machine learning, computer vision and deep learning programs.

Some universities and research groups to find top deep learning scientists include:

Carnegie Mellon, NYU, Stanford, Swiss Al Lab IDSIA, UC Berkeley, University of Michigan, University of Montreal (LISA Lab), University of Oxford, University of Toronto, and several others.

The incredible thing is that the applications of deep learning technologies are expanding quickly with time. It’s certainly slated to play a large role in the tech industry for decades to come.

More like this:

Infographic: 7 Tips to Retain Exceptional Employees

How to Cope with the Stress of a Demanding Job

Top 5 Lessons Companies Can Learn from Burning Man

Use Technology to Manage Your Time Like a Boss