June 16, 2016
by Anand Rao
Some firms are using machine learning to process large amounts of unstructured data, but it’s not widespread—yet.
In the previous post in our Machine Learning series, we dived into the inner workings of deep learning. Given deep learning’s unparalleled power, it’s not surprising that technology companies are competing with one another to collect deep learning experts and apply the techniques to their own operations. How are companies using deep learning to drive business goals?
The tech giants have been using the technique for improved image and video recognition, audio recognition, and language understanding and are actively contributing to open-source research tools. Meanwhile, start-ups are serving boutique needs.
A host of user-friendly libraries and tools have recently emerged that allow quick implementations of deep neural networks. A data scientist can put together a solution through a series of installations and a handful of lines of code. However, a model’s hyper-parameters (how many cycles the neural net goes through, how quickly the net adjusts, etc.) still have to be fitted to the dataset in question to optimize results.
In addition, deep learning is suitable only for certain classes of problems and cannot be seen as a panacea to solve all problems. Deep learning and how it recognizes patterns and outputs its results to a large extent is a ‘black box.’ Areas of investigation where the results of the computation need to be explained to a human are not suitable for this type of learning approach. Also, deep learning requires large volumes of data in order for a system to learn the features and should not be attempted where data is sparse. We would recommend an incremental deployment within enterprises by selecting applications related to large scale image processing. This will enable enterprises to build up the relevant technical capabilities and move through the steep learning curve for this type of learning.
Are you ready to use deep learning techniques in your business? Do you know where it would be most applicable and do you have the necessary data to train the system? Have your data scientists experimented with open source tools currently available? What are some of the challenges you face in implementing and rolling out deep learning systems in your organization?