Deep-Earning, or, Why Attend a Deep-Learning Workshop

Sep 27, 2016   •   Lyndon Maydwell




Kreg Steppe - "Deep in Thought"

Silverpond runs four-hour deep-learning workshops. We want to share our knowledge because machine learning and deep-learning will change how we use software to solve problems. We have a technical understanding of the concepts and mechanisms involved in deep-learning, but we also have a vision for how industry adoption of deep-learning will change the way data scientists and software engineers work in the future.

Why would a business care about deep-learning?

It’s worth considering the following advantages of having an in-house understanding of deep-learning concepts. While your business may not traditionally have played in the machine-learning space, some of the barriers to this space, or motivations for entering have changed recently with the advent of accessible and effective deep-learning techniques.

Hard reasons:

  • Analysis of previously inaccessible data
  • Learned features
  • Human-Level intelligence deployed at scale
  • Reuse / retraining of existing models

Soft reasons:

  • Industry trends
  • Informed Recruitment / HR
  • Informed Management
  • Signalling

Hard reasons

There are strong technical reasons for adopting deep-learning. Advantages in terms of new capabilities and competitive edges spring from increased analytic capacity and new automation opportunities.

Analysis of previously inaccessible data

Until recently, business intelligence, prediction, and recommendations were heavily constrained by available metadata and sophisticated and time-consuming hand-coded heuristics. Deep-learning makes the analysis of new data-types possible, opening the black-boxes of textual, graphical and time-series analysis to the business for profit opportunities and cost-control.

Learned features

The construction of feature-analysis in traditional machine-learning and hand-coding of heuristic methods was time-consuming, required subject-matter-experts, and did not adapt to changes in the world that broke your assumptions. Deep-learning models can mitigate these issues by learning “end-to-end” how to construct conceptual embeddings of your problem domains and how to best leverage this to provide you with predictions and answers. Construction of interoperable networks will even allow you to gain business-domain understanding from the models that you have trained.

Human-level intelligence deployed at scale

Given that deep-learning can enable machines to tackle previously intractable problems that required specialised human analysis, the major implication for a business is that this capability is now scalable. Buy more servers; predict and classify more, faster. Your ability to scale up operations can now be planned in terms of minutes as opposed to the months required for recruitment / training.

Reuse / retraining of existing models

Although deep-learning greatly commoditizes the notion of training through the concept of “end-to-end” learning, you can also take advantage of the great wealth of existing pre-trained networks available for commercial use for free. To foster the ecosystem and promote their use of these techniques, giants like Google and Facebook have made many of their trained models available for use by the public. Your business can take advantage of this by applying the use of these models where they fit exactly, and retraining the models (at greatly reduced cost) where they fit approximately.

Soft reasons

While the technical reasons should be extremely compelling on their own, there are also “softer” reasons why gaining a working understanding of deep-learning may be a good idea for a business.

There is no denying that deep-learning is extremely hot right now and that many companies, large-and-small, are adopting the techniques to extend their capabilities and gain a competitive edge. Even if you do not end up using deep-learning in your business it makes sense to be informed enough so that you can make the decision intelligently rather than defaulting because of a lack of understanding of what deep-learning can and cannot offer you.

Informed Recruitment / HR

If you do decide to invest in a deep-learning capability for your business, then you will undoubtedly need to acquire the skills of talented individuals to aid you in your journey. Either through hiring, or by enlisting the services of consultants. If you want to be able to evaluate the sophistication of these individuals then it makes sense that you should have an understanding, or at least a good intuition for the concepts involved, so that you can evaluate expected performance accurately.

Informed Management

Likewise, when you have built a talented deep-learning team, you should be able to make best use of their efforts, cultivate their growth, and curate the composition of the team to maximise effectiveness. An understanding of deep-learning will also allow you to identify upcoming trends that will allow for strategic research to best plan your strategic roadmap.

Signalling

Finally, by being on the cutting edge of trends such as deep-learning your business signals that it is a savvy operator and is investing in future growth. This can appeal both to customers, who want to ensure that their service-providers will stay relevant in the future, and to investors, who want to foster the growth of businesses who are well placed to take advantage of emerging capabilities.


Come Along!

We’d love to see you at one of our workshops - get your ticket early to secure a place. The workshop provides great value in terms of upskilling on deep-learning concepts. Garner support from your supervisor by sharing this post with them!


Level 2
382 Little Collins Street
Melbourne VIC, 3000

Enter from McKillop Street

Contact us

(03) 9008 5922

Get the latest on machine learning and data science in your inbox each month