A 'deep learning' platform developed by Californian startup Erstaz Labs now provides client companies with simplified solutions for automated learning.

Ersatz Labs Offers 'Deep Learning' Service to Companies

Listed by MIT Tech Review among the top 10 breakthroughs of 2013, deep learning uses algorithms from the field of automated learning or ‘machine learning’ to simulate how neuron activity functions in a person’s brain. The result of this data analysis technique will be deep learning neural networks.  Deep learning is already being applied by major companies including Google and Facebook to such areas as image recognition and predicting people’s online behaviour, and is increasingly being used by all kinds of firms to enhance their overall grasp of their business Now San Francisco-based startup Ersatz Labs has developed the first commercial platform for deep learning, offering its clients a means of using the algorithms to analyse their data, either as a service in the Cloud or via an in-house hardware appliance with installed software.

First-ever commercial deep learning platform

Companies can call on the services of Ersatz Labs to meet a range of data-related needs. They may wish for example to forecast sales figures, predict the click rate on an advertisement, or to assist with voice identification. A client company will then collect historical data related to the target need and upload it to the Ersatz Labs secure Cloud server. After a ‘training’ period during which the neural network ‘learns’ from the company data, Ersatz will be able to make predictions about future results based on the examples shown to the system. The unique feature of the startup’s service offer is that it provides any company interested in deep learning with an easy-to-use web interface – either the Cloud option or in the form of ready-to-use software – without first having to set up complex systems. Ersatz is charging 40 cents a minute of training for the Cloud service. For the basic appliance-plus-software, the company is asking $35,000.

Optimising data use

The recent advent of the Big Data phenomenon has seen the collection of enormous quantities of data.  The purpose of ‘deep learning’, which first emerged as long ago as 2006, is to sift this mass of data in order to extract specific information. When he founded Ersatz Labs in 2012, Dave Sullivan was basically aiming to provide a machine learning capability as a service to companies. This service has been in beta for the past year with 2,200 clients in the fields of marketing, finance and sales, plus also the oil industry. These businesses have a lot to learn from use of the algorithms and from machine learning, with a view to optimising their various business applications.

By Eliane HONG