Intro to deep learning

Intro to deep learning

• 1.
  INTRODUCTION TO DEEP LEARNING @DaveVoyles Sr.Technical Evangelist, Microsoft DaveVoyles.com
• 2.
  WHAT IS DEEPLEARNING? First conceived in the 1950s, it is a class or subset of machine learning algorithms that learns by using a large, many-layered collection of connected processes and exposing these processors to a vast set of examples. One of its primary attributes is the ability to identify patterns in unstructured data.
• 3.
  WHY A SUDDENRESURGENCE? Advanced algorithms are developing as a result of rapid improvements in: • Fast information storage capacity, • High computing power • Parallelization
• 4.
  USE CASES &APPLICATIONS Use Cases: Computer vision, voice recognition, and natural language processing (NLP). Business Applications: Text-based searches, fraud detection, handwriting recognition, image search, and translation. Problems lending themselves to DL include medical diagnosis, demand prediction, malware detection, self- driving cars, customer churn, and failure prediction
• 5.
  USE CASES &APPLICATIONS Use Cases: Computer vision, voice recognition, and natural language processing (NLP). Business Applications: Text-based searches, fraud detection, handwriting recognition, image search, and translation. Problems lending themselves to DL include medical diagnosis, demand prediction, malware detection, self- driving cars, customer churn, and failure prediction
• 6.
  SHORTCOMING Can be expensiveand tricky to set up: requirement of a large amount of data to train neural networks. Still a very immature market, and most organizations lack the necessary data science skills for even simple machine learning solutions. Not clear upfront if deep learning will solve a given problem at all – there is simply no mathematical theory available that indicates if a "good enough" deep learning
• 7.
  SHORTCOMING Can be expensiveand tricky to set up: requirement of a large amount of data to train neural networks. Still a very immature market, and most organizations lack the necessary data science skills for even simple machine learning solutions. Not clear upfront if deep learning will solve a given problem at all – there is simply no mathematical theory available that indicates if a "good enough" deep learning
• 8.
  ML VS DL DLmodel: Able to learn on its own, ML model: Needs to be told how it should make an accurate prediction (by feeding it more data). Conceptually, DL is like ML but different because it can work directly on digital representations of data DL potentially limit human biases that go into choosing inputs, but also find more meaningful measures than the input ML relies on
• 9.
  ALGORITHMS Deep neural networks(DNNs): The dominant deep learning algorithms, which are neural networks constructed from many layers ("deep") of alternating linear & nonlinear processing units Random Decision Forests (RDFs): Also constructed from many layers, but instead of neurons the RDF is constructed from decision trees & outputs a statistical average of the individual trees.
• 10.
  HUMBLE BEGINNINGS
• 11.
  NEURAL NETS First conceivedin the 1950, although many of the key algorithmic advances occurred in the 1980s and 1990s.
• 12.
  BOLTZMANN MACHINE Terry Sejnowskideveloped the basic algorithms called a Boltzmann machine in the early 1980s, which is a network of symmetrically connected, neuron-like units that make stochastic decisions about whether to be on or off.
• 13.
  TERM “DEEP LEARNING” Startedgaining acceptance after a publication by U. of Toronto professor Geoffrey Hinton & grad student Ruslan Salakhutdinov. In 2006 they showed that neural nets could be adequately pre- trained one layer at a time, accelerate consecutive supervised learning, which would then fine- tune the outcome
• 14.
  CURRENT STATE OFTHE MARKET
• 15.
  OPEN SOURCE FRAMEWORKS
• 16.
  HARDWARE NVIDIA: Kepler GPUspowering Microsoft & Amazon's cloud, Jetson TK-x & DGX-1 Microsoft, July 2017: Chip created for HoloLens that includes a module custom-designed to efficiently run deep learning software Google, May 2016: Using its own tailor-made chips called Tensor Processing Units (TPUs) FPGAs (field-programmable gate arrays): Ability to provide a higher performance per watt of power consumption vs GPUs
• 17.
  BIG DEMAND According toMicrosoft CVP Peter Lee, there’s a “bloody war for talent in this space.” Given their size, Google, Facebook, Microsoft, and NVIDIA can afford to hire the most accomplished deep learning talent and pay them handsomely. Deep learning represented almost half of all enterprise AI revenue in 2016, according to Tractica,