1. SAME DATA.BETTER RESULTS.PAUL SALAZARPAUL@SKYTREE.NET!1
2. SKYTREE’S FOCUS"PRODUCTION GRADE"MACHINE LEARNINGMachine learning: the modern science of ﬁnding patterns and making predictions from data.!aka: multivariate statistics, data mining, pattern recognition, or advanced/predictive analytics.!
3. Machine Learning Use Cases!Predict categories and classes!Predict values and numbers!Grouping and segmentation!Detection and characterization!Visualization and reduction!Find similar items !Classiﬁcation !Regression!Clustering!Density Estimation !Dimension Reduction!Multidimensional Querying!Example Skytree Algorithms: Random Decision Forests, Gradient Boosting Machines, NearestNeighbor, Kernel Density Estimation, K-means, Linear Regression, Support Vector Machine,2-point Correlation, Decision Tree, Singular Value Decomposition, Range Search, Logistic RegressionRecommendations PredictionsOutlierDetection
4. What are the current options for ML for Big Data!1. Just use a subset of the data!!– e.g. just take the ﬁrst 1,000 rows. Result to expect: Capture onlythe broadest patterns. à Lower accuracy."2. Just use a simple ML method!!– e.g. use logistic regression instead of nonlinear SVM. Result toexpect: Entire types of patterns cannot be found. à Loweraccuracy."3. Just use simple parallelism/MapReduce!!– i.e. replace all the for-loops with parallel ones. Result to expect:Only the simplest of ML methods (not O(N2)/O(N3)) can besigniﬁcantly sped up this way. à See #2."4. Just throw it in the cloud!!– i.e. somehow use the large compute power of the cloud. Resultto expect: The cost of sending it to the cloud is even greater thanthe compute cost. à See #1. See also #3."
5. Skytree’s Unique Differentiation: Fundamental Technology Breakthrough!Complexity of State-of-the-Art Machine Learning methods:!1. Querying: all-nearest-neighbors O(N2)!2. Density estimation: kernel density estimation O(N2), kernel conditional density est.O(N3) !3. Classiﬁcation: logistic regression, decision tree, neural nets, nearest-neighbor classiﬁer O(N2), kernel discriminant O(N2), support vector machine O(N3), !4. Regression: linear regression, LASSO, kernel regression O(N2), regression tree, Gaussian process regression O(N3)!5. Dimension reduction: PCA, non-negative matrix factorization, kernel PCA O(N3), maximum variance unfolding O(N3); Gaussian graphical models, discrete graphicalmodels!6. Clustering: k-means, mean-shift O(N2), hierarchical clustering O(N3)!7. Testing and matching: MST O(N3), bipartite cross-matching O(N3), n-point correlation 2-sample testing O(Nn), n=2, 3, 4, …!► Unfortunately O(N2), O(N3) are computationally prohibitive for big data!Skytree has invented a way to reduce the complexity of abovemethods from O(N2) and O(N3) to O(N) or O(N log N).5
6. Performance!Up to 10,000x !speedups!(on one CPU)!6
7. How Does Skytree Do This?!7Deep knowledge of algorithmsDrawing from the latest from academiaSmart programmingEfficient ways to compute order N(2) and N(3)Distributed systemsTake advantage of parallel computing speed
8. Team!8Martin Hack, CEO & Co-Founder Sun, GreenBorder (Google)!Alexander Gray, PhD, CTO & Co-Founder Leading Light for Large-Scale, Fast Algorithms!Paul Salazar, VP Sales RedHat, Greenplum!Leland Wilkinson, PhD, VP Data Visualization Creator of SYSTAT (SPSS/IBM).!Tim Marsland, PhD, VP Engineering Sun Fellow, CTO Software, Apple, Oracle!!!!EXECUTIVETEAM!BOARD OFDIRECTORS!Rick Lewis, USVP Noah Doyle, Javelin Venture Partners!David Toth, Founder and CEO NetRatings (Nielsen)!Prof. Michael Jordan, UC Berkeley: machine learning ‘godfather’!Prof. David Patterson, UC Berkeley: systems (inventor RISC, RAID)!Prof. Pat Hanrahan, Stanford: data visualization (Tableau, Pixar)!Prof. James Demmel, UC Berkeley: high-performance computing!INVESTORS!TECH!ADVISORY!BOARD!USVP, Javelin Venture Partners, Scott McNealy, UPS