PPT-Minerva: A Scalable and Highly Efficient Training Platform for Deep Learning

M Wang T Xiao J Li J Zhang C Hong amp Z Zhang 2014 Presentation by Cameron Hamilton Overview Problem disparity between deep learning tools oriented towards productivitygenerality eg MATLAB and taskspecific tools designed for speed and scale eg CUDA. Adam Coates. Stanford University. (Visiting Scholar: Indiana University, Bloomington). What do we want ML to do?. Given image, predict complex high-level patterns:. Object recognition. Detection. Segmentation. Early Work. Why Deep Learning. Stacked Auto Encoders. Deep Belief Networks. CS 678 – Deep Learning. 1. Deep Learning Overview. Train networks with many layers (vs. shallow nets with just a couple of layers). MINERvA. Kevin McFarland. University of Rochester. NuINT11, Dehradun. 8 March 2011. ν. Goddess′. μ. To India . via. Indiana. Thank you to the organizers . for their efforts to allow me to attend . MINERvA is studying neutrino interactions in unprecedented detail on a variety of different nuclei. Low . Energy (LE) . Beam Goals: . S. tudy . both signal and background reactions relevant . to oscillation experiments (current and future). MapReduce. Karim Ibrahim. Anh Pham. 1. Outline. Motivation & Introduction. Benchmarking . study. Optimizing . Hadoop. Motivation. 3. Motivation (. cont. ’). Motivation. :. MapReduce. model is not well suited for one pass analytics.. David . Kaser. Lecture Series. Lorcan Dempsey / . @. LorcanD. Indiana University, . 7 October 2012. How terrific to see you are the featured lecturer this year.   Just thought I'd mention that David . Commutativity. Rule: Designing Scalable Software for Multicore Processors. Austin T. Clements, M. . Frans. . Kaashoek. , . Nickolai. . Zeldovich. , Robert T. Morris, and Eddie Kohler. MIT CSAIL and Harvard University. Processors. Presented by . Remzi. Can . Aksoy. *Some slides . are. . borrowed from a ‘Papers We Love’ . Presentation. EECS 582 – F16. 1. Outline. The . Scalable Commutativity Rule: . Whenever interface operations commute, they can be implemented in a way that scales. Carey . Nachenberg. Deep Learning for Dummies (Like me) – Carey . Nachenberg. (Like me). The Goal of this Talk?. Deep Learning for Dummies (Like me) – Carey . Nachenberg. 2. To provide you with . Luis . Herranz. Arribas. Supervisor: Dr. José M. Martínez Sánchez. Video Processing and Understanding Lab. Universidad . Aut. ónoma. de Madrid. Outline. Introduction. Integrated. . summarization. Aaron Crandall, 2015. What is Deep Learning?. Architectures with more mathematical . transformations from source to target. Sparse representations. Stacking based learning . approaches. Mor. e focus on handling unlabeled data. Topic 3. 4/15/2014. Huy V. Nguyen. 1. outline. Deep learning overview. Deep v. shallow architectures. Representation learning. Breakthroughs. Learning principle: greedy layer-wise training. Tera. . scale: data, model, . Garima Lalwani Karan Ganju Unnat Jain. Today’s takeaways. Bonus RL recap. Functional Approximation. Deep Q Network. Double Deep Q Network. Dueling Networks. Recurrent DQN. Solving “Doom”. Large scale computing systems. Scalability . issues. Low level and high level communication abstractions in scalable systems. Network interface . Common techniques for high performance communication.