2. Blackbox hyperparameter optimization and AutoML

Size: px
Start display at page:

Download "2. Blackbox hyperparameter optimization and AutoML"

Transcription

1 AutoML Automatic Selection, Configuration & Composition of ML Algorithms. at ECML PKDD 2017, Skopje. 2. Blackbox hyperparameter optimization and AutoML Pavel Brazdil, Frank Hutter, Holger Hoos, Joaquin Vanschoren

2 Outline 2 Blackbox hyperparameter optimization AutoML systems based on blackbox hyperparameter optimization

3 Blackbox Hyperoptimization of ML Algorithms 3 Hyperparameter configuration Train & evaluate ML algorithm Cross-validation performance f( ) Hyperparameter optimizer

4 Blackbox Hyperoptimization of ML Pipelines 4 Cross-validation performance f( ) Hyperparameter optimizer

5 A General Notion of Hyperparameters 5 Hyperparameter types Continuous (e.g., learning rate), integer (e.g, #units), ordinal Categorical: finite domain, unordered, e.g. {SVM, RF, NN} Hyperparameter space has structure E.g., top-level hyperparameter A chooses algorithm: {SVM, RF} SVM s soft margin C is only active if A = SVM C is a conditional hyperparameter with parent A Hyperparameters give rise to a structured space of algorithms Many configurations (e.g ) Configurations often yield qualitatively different behaviour

6 The Simplest Strategy: Random Search 6 Select configurations uniformly at random Completely uninformed Global search, won t get stuck in a local region At least it s better than grid search: Image source: Bergstra et al, Random Search for Hyperparameter Optimization, JMLR 2012

7 7 The Other Extreme: Gradient Descent (aka hill climbing) Start with some configuration repeat Modify a single parameter if performance on a benchmark set degrades then undo modification until no more improvement possible (or good enough")

8 Stochastic Local Search 8 Balance intensification and diversification [e.g., Hoos and Stützle, 2005] Intensification: gradient descent Diversification: restarts, random steps, perturbations, Prominent general methods Tabu search [Glover, 1986] Simulated annealing [Kirkpatrick, Gelatt, C. D.; Vecchi, 1983] Iterated local search [Lourenço, Martin & Stützle, 2003]

9 Population-based Methods 9 Population of configurations Global + local search via population Maintain population fitness & diversity Examples Genetic algorithms [e.g., Barricelli, 57, Goldberg, 89] Evolutionary strategies [e.g., Beyer & Schwefel, 02]

10 Estimation of Distribution (EDA) 10 [e.g., Pelikan, Goldberg and Lobo, 2002] Categorize performance into good and bad, and fit a model (density estimator) of the good points in the space: P(x is good ) Often: independent Gaussians for each dimension Sample next point to evaluate from the model Image source: Wikipedia

11 11 Bayesian Optimization Fit a (probabilistic) model of the function p(f x) Use that model to trade off exploitation vs exploration Prominent method for expensive blackbox optimization [Mockus et al., '78] Recent convergence results [Srinivas et al, '10; Bull '11; de Freitas, Smola, Zoghi, '12, Kawaguchi et al, '15]

12 AutoML Challenges for Bayesian Optimization 12 Problems for standard Gaussian Process (GP) approach: Complex hyperparameter space High-dimensional (low effective dimensionality) Mixed continuous/discrete hyperparameters Conditional hyperparameters Discrete change points Noise: sometimes heteroscedastic, large, non-gaussian Robustness (usability out of the box) Model overhead (budget is runtime, not #function evaluations) Simple solution: random forests [Breiman, '01] Adapted to yield uncertainty estimates as a mixture model over trees

13 13 Bayesian Optimization with Random Forests [Hutter, Hoos, Leyton-Brown, ] SMAC: Sequential Model-Based Algorithm Configuration repeat construct RF model to predict performance use that model to select promising configurations compare each selected configuration against the best known until time budget exhausted Distributed SMAC Maintain queue of promising configurations Compare these to * on distributed worker cores

14 Comparing Bayesian Hyperparameter Optimizers 14 [Eggensperger, Feurer, Hutter, Bergstra, Snoek, Hoos & Leyton-Brown, BayesOpt 2013] Hyperparameter optimization library: automl.org/hpolib Benchmarks From 2-dimensional continuous hyperparameter spaces To structured ones with 768 hyperparameters Optimizers SMAC [Hutter et al, '11], based on random forests Spearmint [Snoek et al, '12], based on Gaussian processes TPE [Bergstra et al, '11], based on 1-d distributions of good values Results GP-based Spearmint is best for low-dimensional & continuous RF-based SMAC is best for high-dim, categorical & conditional

15 Neural networks to the rescue? 15 Two recent promising models for Bayesian optimization Neural networks with Bayesian linear regression using the features in the output layer [Snoek et al, ICML 2015] Fully Bayesian neural networks, trained with stochastic gradient Hamiltonian Monte Carlo [Springenberg et al, NIPS 2016] Good performance on low-dimensional HPOlib tasks So far not studied for: High dimensionality Conditional hyperparameters

16 Outline 16 Blackbox hyperparameter optimization AutoML systems based on blackbox hyperparameter optimization

17 Auto-WEKA s AutoML approach 17 Expose the choices in a machine learning framework Algorithms, hyperparameters, preprocessors, Optimize CV performance using SMAC Obtain a true push-button solution for machine learning [Thornton, Hutter, Hoos, Leyton-Brown; KDD 2013] Learn Here: use the broad range of methods implemented in WEKA [Witten et al, 1999-current] 27 base classifiers (with up to 10 parameters each) 10 meta-methods 2 ensemble methods

18 WEKA s configuration space 18 Base classifiers 27 choices, each with up to 10 subparameters Coarse discretization: about 10 8 instantiations Hierarchical structure on top of base classifiers

19 WEKA s configuration space (cont d) 19 Feature selection Search method: which feature subsets to evaluate Evaluation method: how to evaluate feature subsets in search Both methods have subparameters about 10 7 instantiations In total: 768 parameters, configurations

20 Auto-WEKA: results 20 Auto-WEKA performed better than best base classifier Even when best base classifier determined by an oracle In 6/21 datasets more than 10% reductions in relative error In WEKA Comparison packageto manager; full grid downloaded search 400 times per week Union of grids over parameters of all 27 base classifiers Auto-WEKA was 100 times faster Auto-WEKA had better test performance in 15/21 cases Auto-WEKA based on SMAC vs. TPE [Bergstra et al, NIPS'11] SMAC yielded better CV performance in 19/21 cases SMAC yielded better test performance in 14/21 cases Differences usually small, in 3 cases substantial (SMAC better)

21 The Auto-WEKA approach applied to deep nets Deep network structure & hyperparameters Cross-validation performance f( ) Bayesian optimization Units per layer dog cat Kernel size # convolutional layers # fully connected layers + Learning rates, batch sizes, dropout rates, 21

22 Application 1: Object Recognition 22 Parameterized the Caffe framework [Jia, 2013] Convolutional neural network with up to 6 layers 81 hyperparameters 9 network hyperparameters 12 layer-wise hyperparameters for each of the 6 layers Results for CIFAR-10 New best result for CIFAR-10 without data augmentation SMAC outperformed TPE (only other applicable hyperparameter optimizer) [Domhan, Springenberg, Hutter, IJCAI 2015]

23 Application 2: Movement Decoding from EEG 23 [Schirrmeister, Fiederer, Springenberg, Eggensperger, Ball, Hutter, Tangermann, Human-Brain Mapping 2017] Convolutional neural network for motor-execution data Tap fingers on left hand / right hand / do nothing / clench toes EEG data from 128 channels Results for Auto-Net Automatically selected useful subset of channels Outperformed manual solution, by 10% relative error Per-patient optimization: cross-validation error rates reduced by factor of 2

24 24 Application 3: AutoML Challenge [Mendoza, Klein, Feurer, Springenberg, Hutter, AutoML 2016] Unstructured data fully-connected network Up to 5 layers (with 3 layer hyperparameters each) 14 network hyperparameters, in total 29 hyperparameters Optimized for 18h on 5GPUs Timeout of 30 minutes per network ( 500 networks evaluated) Auto-Net won several datasets against human experts E.g., Alexis data set: data points, 5000 features, 18 classes Test set AUC 90% All other (manual) approaches < 80% First automated deep learning system to win a ML competition data set against human experts

Automatic Machine Learning (AutoML): A Tutorial

Automatic Machine Learning (AutoML): A Tutorial Automatic Machine Learning (AutoML): A Tutorial Frank Hutter University of Freiburg fh@cs.uni-freiburg.de Joaquin Vanschoren Eindhoven University of Technology j.vanschoren@tue.nl Slides available at automl.org/events

More information

7. Metalearning for Automated Workflow Design

7. Metalearning for Automated Workflow Design AutoML 2017. at ECML PKDD 2017, Skopje. Automatic Selection, Configuration & Composition of ML Algorithms 7. Metalearning for Automated Workflow Design by. Pavel Brazdil, Frank Hutter, Holger Hoos, Joaquin

More information

Overview on Automatic Tuning of Hyperparameters

Overview on Automatic Tuning of Hyperparameters Overview on Automatic Tuning of Hyperparameters Alexander Fonarev http://newo.su 20.02.2016 Outline Introduction to the problem and examples Introduction to Bayesian optimization Overview of surrogate

More information

arxiv: v1 [cs.lg] 23 Oct 2018

arxiv: v1 [cs.lg] 23 Oct 2018 ICML 2018 AutoML Workshop for Machine Learning Pipelines arxiv:1810.09942v1 [cs.lg] 23 Oct 2018 Brandon Schoenfeld Christophe Giraud-Carrier Mason Poggemann Jarom Christensen Kevin Seppi Department of

More information

Bayesian Optimization Nando de Freitas

Bayesian Optimization Nando de Freitas Bayesian Optimization Nando de Freitas Eric Brochu, Ruben Martinez-Cantin, Matt Hoffman, Ziyu Wang, More resources This talk follows the presentation in the following review paper available fro Oxford

More information

Efficient Hyperparameter Optimization of Deep Learning Algorithms Using Deterministic RBF Surrogates

Efficient Hyperparameter Optimization of Deep Learning Algorithms Using Deterministic RBF Surrogates Efficient Hyperparameter Optimization of Deep Learning Algorithms Using Deterministic RBF Surrogates Ilija Ilievski Graduate School for Integrative Sciences and Engineering National University of Singapore

More information

Learning to Transfer Initializations for Bayesian Hyperparameter Optimization

Learning to Transfer Initializations for Bayesian Hyperparameter Optimization Learning to Transfer Initializations for Bayesian Hyperparameter Optimization Jungtaek Kim, Saehoon Kim, and Seungjin Choi Department of Computer Science and Engineering Pohang University of Science and

More information

CS 179 Lecture 16. Logistic Regression & Parallel SGD

CS 179 Lecture 16. Logistic Regression & Parallel SGD CS 179 Lecture 16 Logistic Regression & Parallel SGD 1 Outline logistic regression (stochastic) gradient descent parallelizing SGD for neural nets (with emphasis on Google s distributed neural net implementation)

More information

BAYESIAN GLOBAL OPTIMIZATION

BAYESIAN GLOBAL OPTIMIZATION BAYESIAN GLOBAL OPTIMIZATION Using Optimal Learning to Tune Deep Learning Pipelines Scott Clark scott@sigopt.com OUTLINE 1. Why is Tuning AI Models Hard? 2. Comparison of Tuning Methods 3. Bayesian Global

More information

Towards efficient Bayesian Optimization for Big Data

Towards efficient Bayesian Optimization for Big Data Towards efficient Bayesian Optimization for Big Data Aaron Klein 1 Simon Bartels Stefan Falkner 1 Philipp Hennig Frank Hutter 1 1 Department of Computer Science University of Freiburg, Germany {kleinaa,sfalkner,fh}@cs.uni-freiburg.de

More information

Sequential Model-based Optimization for General Algorithm Configuration

Sequential Model-based Optimization for General Algorithm Configuration Sequential Model-based Optimization for General Algorithm Configuration Frank Hutter, Holger Hoos, Kevin Leyton-Brown University of British Columbia LION 5, Rome January 18, 2011 Motivation Most optimization

More information

Combining Hyperband and Bayesian Optimization

Combining Hyperband and Bayesian Optimization Combining Hyperband and Bayesian Optimization Stefan Falkner Aaron Klein Frank Hutter Department of Computer Science University of Freiburg {sfalkner, kleinaa, fh}@cs.uni-freiburg.de Abstract Proper hyperparameter

More information

ImageNet Classification with Deep Convolutional Neural Networks

ImageNet Classification with Deep Convolutional Neural Networks ImageNet Classification with Deep Convolutional Neural Networks Alex Krizhevsky Ilya Sutskever Geoffrey Hinton University of Toronto Canada Paper with same name to appear in NIPS 2012 Main idea Architecture

More information

Efficient Hyper-parameter Optimization for NLP Applications

Efficient Hyper-parameter Optimization for NLP Applications Efficient Hyper-parameter Optimization for NLP Applications Lidan Wang 1, Minwei Feng 1, Bowen Zhou 1, Bing Xiang 1, Sridhar Mahadevan 2,1 1 IBM Watson, T. J. Watson Research Center, NY 2 College of Information

More information

Supplementary material for: BO-HB: Robust and Efficient Hyperparameter Optimization at Scale

Supplementary material for: BO-HB: Robust and Efficient Hyperparameter Optimization at Scale Supplementary material for: BO-: Robust and Efficient Hyperparameter Optimization at Scale Stefan Falkner 1 Aaron Klein 1 Frank Hutter 1 A. Available Software To promote reproducible science and enable

More information

Black-Box Hyperparameter Optimization for Nuclei Segmentation in Prostate Tissue Images

Black-Box Hyperparameter Optimization for Nuclei Segmentation in Prostate Tissue Images Black-Box Hyperparameter Optimization for Nuclei Segmentation in Prostate Tissue Images Thomas Wollmann 1, Patrick Bernhard 1,ManuelGunkel 2, Delia M. Braun 3, Jan Meiners 4, Ronald Simon 4, Guido Sauter

More information

HYPERBAND: BANDIT-BASED CONFIGURATION EVAL-

HYPERBAND: BANDIT-BASED CONFIGURATION EVAL- HYPERBAND: BANDIT-BASED CONFIGURATION EVAL- UATION FOR HYPERPARAMETER OPTIMIZATION Lisha Li UCLA lishal@cs.ucla.edu Kevin Jamieson UC Berkeley kjamieson@berkeley.edu Giulia DeSalvo NYU desalvo@cims.nyu.edu

More information

hyperspace: Automated Optimization of Complex Processing Pipelines for pyspace

hyperspace: Automated Optimization of Complex Processing Pipelines for pyspace hyperspace: Automated Optimization of Complex Processing Pipelines for pyspace Torben Hansing, Mario Michael Krell, Frank Kirchner Robotics Research Group University of Bremen Robert-Hooke-Str. 1, 28359

More information

Efficient Benchmarking of Hyperparameter Optimizers via Surrogates

Efficient Benchmarking of Hyperparameter Optimizers via Surrogates Efficient Benchmarking of Hyperparameter Optimizers via Surrogates Katharina Eggensperger and Frank Hutter University of Freiburg {eggenspk, fh}@cs.uni-freiburg.de Holger H. Hoos and Kevin Leyton-Brown

More information

arxiv: v1 [cs.lg] 29 May 2014

arxiv: v1 [cs.lg] 29 May 2014 BayesOpt: A Bayesian Optimization Library for Nonlinear Optimization, Experimental Design and Bandits arxiv:1405.7430v1 [cs.lg] 29 May 2014 Ruben Martinez-Cantin rmcantin@unizar.es May 30, 2014 Abstract

More information

n Informally: n How to form solutions n How to traverse the search space n Systematic: guarantee completeness

n Informally: n How to form solutions n How to traverse the search space n Systematic: guarantee completeness Advanced Search Applications: Combinatorial Optimization Scheduling Algorithms: Stochastic Local Search and others Analyses: Phase transitions, structural analysis, statistical models Combinatorial Problems

More information

Auto-WEKA: Combined Selection and Hyperparameter Optimization of Classification Algorithms

Auto-WEKA: Combined Selection and Hyperparameter Optimization of Classification Algorithms Auto-WEKA: Combined Selection and Hyperparameter Optimization of Classification Algorithms Chris Thornton Frank Hutter Holger H. Hoos Kevin Leyton-Brown Department of Computer Science, University of British

More information

Auto-WEKA: Combined Selection and Hyperparameter Optimization of Supervised Machine Learning Algorithms

Auto-WEKA: Combined Selection and Hyperparameter Optimization of Supervised Machine Learning Algorithms Auto-WEKA: Combined Selection and Hyperparameter Optimization of Supervised Machine Learning Algorithms by Chris Thornton B.Sc, University of Calgary, 2011 a thesis submitted in partial fulfillment of

More information

An Empirical Study of Hyperparameter Importance Across Datasets

An Empirical Study of Hyperparameter Importance Across Datasets An Empirical Study of Hyperparameter Importance Across Datasets Jan N. van Rijn and Frank Hutter University of Freiburg, Germany {vanrijn,fh}@cs.uni-freiburg.de Abstract. With the advent of automated machine

More information

Speeding up Automatic Hyperparameter Optimization of Deep Neural Networks by Extrapolation of Learning Curves

Speeding up Automatic Hyperparameter Optimization of Deep Neural Networks by Extrapolation of Learning Curves Speeding up Automatic Hyperparameter Optimization of Deep Neural Networks by Extrapolation of Learning Curves Tobias Domhan, Jost Tobias Springenberg, Frank Hutter University of Freiburg Freiburg, Germany

More information

Dropout. Sargur N. Srihari This is part of lecture slides on Deep Learning:

Dropout. Sargur N. Srihari This is part of lecture slides on Deep Learning: Dropout Sargur N. srihari@buffalo.edu This is part of lecture slides on Deep Learning: http://www.cedar.buffalo.edu/~srihari/cse676 1 Regularization Strategies 1. Parameter Norm Penalties 2. Norm Penalties

More information

Performance Prediction and Automated Tuning of Randomized and Parametric Algorithms

Performance Prediction and Automated Tuning of Randomized and Parametric Algorithms Performance Prediction and Automated Tuning of Randomized and Parametric Algorithms Frank Hutter 1, Youssef Hamadi 2, Holger Hoos 1, and Kevin Leyton-Brown 1 1 University of British Columbia, Vancouver,

More information

Automatic solver configuration using SMAC

Automatic solver configuration using SMAC Automatic solver configuration using SMAC How to boost performance of your SAT solver? Marius Lindauer 1 University of Freiburg SAT Industrial Day 2016, Bordeaux 1 Thanks to Frank Hutter! Ever looked into

More information

Unsupervised Learning: Clustering

Unsupervised Learning: Clustering Unsupervised Learning: Clustering Vibhav Gogate The University of Texas at Dallas Slides adapted from Carlos Guestrin, Dan Klein & Luke Zettlemoyer Machine Learning Supervised Learning Unsupervised Learning

More information

Tracking Algorithms. Lecture16: Visual Tracking I. Probabilistic Tracking. Joint Probability and Graphical Model. Deterministic methods

Tracking Algorithms. Lecture16: Visual Tracking I. Probabilistic Tracking. Joint Probability and Graphical Model. Deterministic methods Tracking Algorithms CSED441:Introduction to Computer Vision (2017F) Lecture16: Visual Tracking I Bohyung Han CSE, POSTECH bhhan@postech.ac.kr Deterministic methods Given input video and current state,

More information

Supplementary material for the paper: Bayesian Optimization with Robust Bayesian Neural Networks

Supplementary material for the paper: Bayesian Optimization with Robust Bayesian Neural Networks Supplementary material for the paper: Bayesian Optimization with Robust Bayesian Neural Networks Jost Tobias Springenberg Aaron Klein Stefan Falkner Frank Hutter Department of Computer Science University

More information

Automatic Algorithm Configuration based on Local Search

Automatic Algorithm Configuration based on Local Search Automatic Algorithm Configuration based on Local Search Frank Hutter 1 Holger Hoos 1 Thomas Stützle 2 1 Department of Computer Science University of British Columbia Canada 2 IRIDIA Université Libre de

More information

Programming by Optimisation:

Programming by Optimisation: Programming by Optimisation: A Practical Paradigm for Computer-Aided Algorithm Design Holger H. Hoos & Frank Hutter Department of Computer Science University of British Columbia Canada Department of Computer

More information

Applying Supervised Learning

Applying Supervised Learning Applying Supervised Learning When to Consider Supervised Learning A supervised learning algorithm takes a known set of input data (the training set) and known responses to the data (output), and trains

More information

AUTOSTACKER: AN AUTOMATIC EVOLUTIONARY HI-

AUTOSTACKER: AN AUTOMATIC EVOLUTIONARY HI- AUTOSTACKER: AN AUTOMATIC EVOLUTIONARY HI- ERARCHICAL MACHINE LEARNING SYSTEM Anonymous authors Paper under double-blind review ABSTRACT This work provides an automatic machine learning (AutoML) modelling

More information

SpySMAC: Automated Configuration and Performance Analysis of SAT Solvers

SpySMAC: Automated Configuration and Performance Analysis of SAT Solvers SpySMAC: Automated Configuration and Performance Analysis of SAT Solvers Stefan Falkner, Marius Lindauer, and Frank Hutter University of Freiburg {sfalkner,lindauer,fh}@cs.uni-freiburg.de Abstract. Most

More information

ECS289: Scalable Machine Learning

ECS289: Scalable Machine Learning ECS289: Scalable Machine Learning Cho-Jui Hsieh UC Davis Sept 22, 2016 Course Information Website: http://www.stat.ucdavis.edu/~chohsieh/teaching/ ECS289G_Fall2016/main.html My office: Mathematical Sciences

More information

PSU Student Research Symposium 2017 Bayesian Optimization for Refining Object Proposals, with an Application to Pedestrian Detection Anthony D.

PSU Student Research Symposium 2017 Bayesian Optimization for Refining Object Proposals, with an Application to Pedestrian Detection Anthony D. PSU Student Research Symposium 2017 Bayesian Optimization for Refining Object Proposals, with an Application to Pedestrian Detection Anthony D. Rhodes 5/10/17 What is Machine Learning? Machine learning

More information

Perceptron: This is convolution!

Perceptron: This is convolution! Perceptron: This is convolution! v v v Shared weights v Filter = local perceptron. Also called kernel. By pooling responses at different locations, we gain robustness to the exact spatial location of image

More information

Problem Solving and Search in Artificial Intelligence

Problem Solving and Search in Artificial Intelligence Problem Solving and Search in Artificial Intelligence Algorithm Configuration Nysret Musliu Database and Artificial Intelligence Group, Institut of Logic and Computation, TU Wien Motivation Metaheuristic

More information

Automatic Algorithm Configuration based on Local Search

Automatic Algorithm Configuration based on Local Search Automatic Algorithm Configuration based on Local Search Frank Hutter 1 Holger Hoos 1 Thomas Stützle 2 1 Department of Computer Science University of British Columbia Canada 2 IRIDIA Université Libre de

More information

CS 229 Midterm Review

CS 229 Midterm Review CS 229 Midterm Review Course Staff Fall 2018 11/2/2018 Outline Today: SVMs Kernels Tree Ensembles EM Algorithm / Mixture Models [ Focus on building intuition, less so on solving specific problems. Ask

More information

Automated Configuration of MIP solvers

Automated Configuration of MIP solvers Automated Configuration of MIP solvers Frank Hutter, Holger Hoos, and Kevin Leyton-Brown Department of Computer Science University of British Columbia Vancouver, Canada {hutter,hoos,kevinlb}@cs.ubc.ca

More information

Machine Learning. Topic 5: Linear Discriminants. Bryan Pardo, EECS 349 Machine Learning, 2013

Machine Learning. Topic 5: Linear Discriminants. Bryan Pardo, EECS 349 Machine Learning, 2013 Machine Learning Topic 5: Linear Discriminants Bryan Pardo, EECS 349 Machine Learning, 2013 Thanks to Mark Cartwright for his extensive contributions to these slides Thanks to Alpaydin, Bishop, and Duda/Hart/Stork

More information

Metric Learning for Large-Scale Image Classification:

Metric Learning for Large-Scale Image Classification: Metric Learning for Large-Scale Image Classification: Generalizing to New Classes at Near-Zero Cost Florent Perronnin 1 work published at ECCV 2012 with: Thomas Mensink 1,2 Jakob Verbeek 2 Gabriela Csurka

More information

What is machine learning?

What is machine learning? Machine learning, pattern recognition and statistical data modelling Lecture 12. The last lecture Coryn Bailer-Jones 1 What is machine learning? Data description and interpretation finding simpler relationship

More information

Local Search (Greedy Descent): Maintain an assignment of a value to each variable. Repeat:

Local Search (Greedy Descent): Maintain an assignment of a value to each variable. Repeat: Local Search Local Search (Greedy Descent): Maintain an assignment of a value to each variable. Repeat: Select a variable to change Select a new value for that variable Until a satisfying assignment is

More information

Applied Statistics for Neuroscientists Part IIa: Machine Learning

Applied Statistics for Neuroscientists Part IIa: Machine Learning Applied Statistics for Neuroscientists Part IIa: Machine Learning Dr. Seyed-Ahmad Ahmadi 04.04.2017 16.11.2017 Outline Machine Learning Difference between statistics and machine learning Modeling the problem

More information

CS839: Probabilistic Graphical Models. Lecture 10: Learning with Partially Observed Data. Theo Rekatsinas

CS839: Probabilistic Graphical Models. Lecture 10: Learning with Partially Observed Data. Theo Rekatsinas CS839: Probabilistic Graphical Models Lecture 10: Learning with Partially Observed Data Theo Rekatsinas 1 Partially Observed GMs Speech recognition 2 Partially Observed GMs Evolution 3 Partially Observed

More information

Machine Learning. Deep Learning. Eric Xing (and Pengtao Xie) , Fall Lecture 8, October 6, Eric CMU,

Machine Learning. Deep Learning. Eric Xing (and Pengtao Xie) , Fall Lecture 8, October 6, Eric CMU, Machine Learning 10-701, Fall 2015 Deep Learning Eric Xing (and Pengtao Xie) Lecture 8, October 6, 2015 Eric Xing @ CMU, 2015 1 A perennial challenge in computer vision: feature engineering SIFT Spin image

More information

An Empirical Study of Per-Instance Algorithm Scheduling

An Empirical Study of Per-Instance Algorithm Scheduling An Empirical Study of Per-Instance Algorithm Scheduling Marius Lindauer, Rolf-David Bergdoll, and Frank Hutter University of Freiburg Abstract. Algorithm selection is a prominent approach to improve a

More information

Machine Learning. The Breadth of ML Neural Networks & Deep Learning. Marc Toussaint. Duy Nguyen-Tuong. University of Stuttgart

Machine Learning. The Breadth of ML Neural Networks & Deep Learning. Marc Toussaint. Duy Nguyen-Tuong. University of Stuttgart Machine Learning The Breadth of ML Neural Networks & Deep Learning Marc Toussaint University of Stuttgart Duy Nguyen-Tuong Bosch Center for Artificial Intelligence Summer 2017 Neural Networks Consider

More information

Neural Networks. CE-725: Statistical Pattern Recognition Sharif University of Technology Spring Soleymani

Neural Networks. CE-725: Statistical Pattern Recognition Sharif University of Technology Spring Soleymani Neural Networks CE-725: Statistical Pattern Recognition Sharif University of Technology Spring 2013 Soleymani Outline Biological and artificial neural networks Feed-forward neural networks Single layer

More information

Bayesian model ensembling using meta-trained recurrent neural networks

Bayesian model ensembling using meta-trained recurrent neural networks Bayesian model ensembling using meta-trained recurrent neural networks Luca Ambrogioni l.ambrogioni@donders.ru.nl Umut Güçlü u.guclu@donders.ru.nl Yağmur Güçlütürk y.gucluturk@donders.ru.nl Julia Berezutskaya

More information

Neural Networks. Single-layer neural network. CSE 446: Machine Learning Emily Fox University of Washington March 10, /10/2017

Neural Networks. Single-layer neural network. CSE 446: Machine Learning Emily Fox University of Washington March 10, /10/2017 3/0/207 Neural Networks Emily Fox University of Washington March 0, 207 Slides adapted from Ali Farhadi (via Carlos Guestrin and Luke Zettlemoyer) Single-layer neural network 3/0/207 Perceptron as a neural

More information

arxiv: v3 [cs.lg] 24 Jun 2016

arxiv: v3 [cs.lg] 24 Jun 2016 FLASH: Fast Bayesian Optimization for Data Analytic Pipelines arxiv:1602.06468v3 [cs.lg] 24 Jun 2016 Yuyu Zhang Mohammad Taha Bahadori Hang Su Jimeng Sun Georgia Institute of Technology {yuyu,bahadori,hangsu}@gatech.edu,jsun@cc.gatech.edu

More information

Neural Networks and Deep Learning

Neural Networks and Deep Learning Neural Networks and Deep Learning Example Learning Problem Example Learning Problem Celebrity Faces in the Wild Machine Learning Pipeline Raw data Feature extract. Feature computation Inference: prediction,

More information

P4ML: A Phased Performance-Based Pipeline Planner for Automated Machine Learning

P4ML: A Phased Performance-Based Pipeline Planner for Automated Machine Learning Proceedings of Machine Learning Research 1:1 8, 2018 ICML2018AutoMLWorkshop P4ML: A Phased Performance-Based Pipeline Planner for Automated Machine Learning Yolanda Gil, Ke-Thia Yao, Varun Ratnakar, Daniel

More information

Deep Learning for Computer Vision

Deep Learning for Computer Vision Deep Learning for Computer Vision Lecture 7: Universal Approximation Theorem, More Hidden Units, Multi-Class Classifiers, Softmax, and Regularization Peter Belhumeur Computer Science Columbia University

More information

Lecture #11: The Perceptron

Lecture #11: The Perceptron Lecture #11: The Perceptron Mat Kallada STAT2450 - Introduction to Data Mining Outline for Today Welcome back! Assignment 3 The Perceptron Learning Method Perceptron Learning Rule Assignment 3 Will be

More information

AI-Augmented Algorithms

AI-Augmented Algorithms AI-Augmented Algorithms How I Learned to Stop Worrying and Love Choice Lars Kotthoff University of Wyoming larsko@uwyo.edu Warsaw, 17 April 2019 Outline Big Picture Motivation Choosing Algorithms Tuning

More information

BOHB: Robust and Efficient Hyperparameter Optimization at Scale

BOHB: Robust and Efficient Hyperparameter Optimization at Scale Stefan Falkner 1 Aaron Klein 1 Frank Hutter 1 Abstract Modern deep learning methods are very sensitive to many hyperparameters, and, due to the long training times of state-of-the-art models, vanilla Bayesian

More information

Tutorial on Machine Learning Tools

Tutorial on Machine Learning Tools Tutorial on Machine Learning Tools Yanbing Xue Milos Hauskrecht Why do we need these tools? Widely deployed classical models No need to code from scratch Easy-to-use GUI Outline Matlab Apps Weka 3 UI TensorFlow

More information

Structure Optimization for Deep Multimodal Fusion Networks using Graph-Induced Kernels

Structure Optimization for Deep Multimodal Fusion Networks using Graph-Induced Kernels Structure Optimization for Deep Multimodal Fusion Networks using Graph-Induced Kernels Dhanesh Ramachandram 1, Michal Lisicki 1, Timothy J. Shields, Mohamed R. Amer and Graham W. Taylor1 1- Machine Learning

More information

Feature Selection. CE-725: Statistical Pattern Recognition Sharif University of Technology Spring Soleymani

Feature Selection. CE-725: Statistical Pattern Recognition Sharif University of Technology Spring Soleymani Feature Selection CE-725: Statistical Pattern Recognition Sharif University of Technology Spring 2013 Soleymani Outline Dimensionality reduction Feature selection vs. feature extraction Filter univariate

More information

Using Machine Learning to Identify Security Issues in Open-Source Libraries. Asankhaya Sharma Yaqin Zhou SourceClear

Using Machine Learning to Identify Security Issues in Open-Source Libraries. Asankhaya Sharma Yaqin Zhou SourceClear Using Machine Learning to Identify Security Issues in Open-Source Libraries Asankhaya Sharma Yaqin Zhou SourceClear Outline - Overview of problem space Unidentified security issues How Machine Learning

More information

Scalable Meta-Learning for Bayesian Optimization using Ranking-Weighted Gaussian Process Ensembles

Scalable Meta-Learning for Bayesian Optimization using Ranking-Weighted Gaussian Process Ensembles ICML 2018 AutoML Workshop Scalable Meta-Learning for Bayesian Optimization using Ranking-Weighted Gaussian Process Ensembles Matthias Feurer University of Freiburg Benjamin Letham Facebook Eytan Bakshy

More information

10703 Deep Reinforcement Learning and Control

10703 Deep Reinforcement Learning and Control 10703 Deep Reinforcement Learning and Control Russ Salakhutdinov Machine Learning Department rsalakhu@cs.cmu.edu Policy Gradient I Used Materials Disclaimer: Much of the material and slides for this lecture

More information

Facial Expression Classification with Random Filters Feature Extraction

Facial Expression Classification with Random Filters Feature Extraction Facial Expression Classification with Random Filters Feature Extraction Mengye Ren Facial Monkey mren@cs.toronto.edu Zhi Hao Luo It s Me lzh@cs.toronto.edu I. ABSTRACT In our work, we attempted to tackle

More information

Gaussian Processes for Robotics. McGill COMP 765 Oct 24 th, 2017

Gaussian Processes for Robotics. McGill COMP 765 Oct 24 th, 2017 Gaussian Processes for Robotics McGill COMP 765 Oct 24 th, 2017 A robot must learn Modeling the environment is sometimes an end goal: Space exploration Disaster recovery Environmental monitoring Other

More information

Data Mining. Introduction. Hamid Beigy. Sharif University of Technology. Fall 1395

Data Mining. Introduction. Hamid Beigy. Sharif University of Technology. Fall 1395 Data Mining Introduction Hamid Beigy Sharif University of Technology Fall 1395 Hamid Beigy (Sharif University of Technology) Data Mining Fall 1395 1 / 21 Table of contents 1 Introduction 2 Data mining

More information

Hill Climbing. Assume a heuristic value for each assignment of values to all variables. Maintain an assignment of a value to each variable.

Hill Climbing. Assume a heuristic value for each assignment of values to all variables. Maintain an assignment of a value to each variable. Hill Climbing Many search spaces are too big for systematic search. A useful method in practice for some consistency and optimization problems is hill climbing: Assume a heuristic value for each assignment

More information

9. Conclusions. 9.1 Definition KDD

9. Conclusions. 9.1 Definition KDD 9. Conclusions Contents of this Chapter 9.1 Course review 9.2 State-of-the-art in KDD 9.3 KDD challenges SFU, CMPT 740, 03-3, Martin Ester 419 9.1 Definition KDD [Fayyad, Piatetsky-Shapiro & Smyth 96]

More information

A Brief Look at Optimization

A Brief Look at Optimization A Brief Look at Optimization CSC 412/2506 Tutorial David Madras January 18, 2018 Slides adapted from last year s version Overview Introduction Classes of optimization problems Linear programming Steepest

More information

Fast Bayesian Optimization of Machine Learning Hyperparameters on Large Datasets

Fast Bayesian Optimization of Machine Learning Hyperparameters on Large Datasets Fast Bayesian Optimization of Machine Learning Hyperparameters on Large Datasets Aaron Klein 1 Stefan Falkner 1 Simon Bartels 2 Philipp Hennig 2 Frank Hutter 1 1 {kleinaa, sfalkner, fh}@cs.uni-freiburg.de

More information

AlphaD3M: Machine Learning Pipeline Synthesis

AlphaD3M: Machine Learning Pipeline Synthesis ICML 2018 AutoML Workshop AlphaD3M: Machine Learning Pipeline Synthesis Iddo Drori idrori@nyu.edu Yamuna Krishnamurthy yamuna@nyu.edu Remi Rampin remi.rampin@nyu.edu Raoni de Paula Lourenco raoni@nyu.edu

More information

Data Mining Practical Machine Learning Tools and Techniques. Slides for Chapter 6 of Data Mining by I. H. Witten and E. Frank

Data Mining Practical Machine Learning Tools and Techniques. Slides for Chapter 6 of Data Mining by I. H. Witten and E. Frank Data Mining Practical Machine Learning Tools and Techniques Slides for Chapter 6 of Data Mining by I. H. Witten and E. Frank Implementation: Real machine learning schemes Decision trees Classification

More information

Deep Learning with Tensorflow AlexNet

Deep Learning with Tensorflow   AlexNet Machine Learning and Computer Vision Group Deep Learning with Tensorflow http://cvml.ist.ac.at/courses/dlwt_w17/ AlexNet Krizhevsky, Alex, Ilya Sutskever, and Geoffrey E. Hinton, "Imagenet classification

More information

How Learning Differs from Optimization. Sargur N. Srihari

How Learning Differs from Optimization. Sargur N. Srihari How Learning Differs from Optimization Sargur N. srihari@cedar.buffalo.edu 1 Topics in Optimization Optimization for Training Deep Models: Overview How learning differs from optimization Risk, empirical

More information

Uninformed Search Methods. Informed Search Methods. Midterm Exam 3/13/18. Thursday, March 15, 7:30 9:30 p.m. room 125 Ag Hall

Uninformed Search Methods. Informed Search Methods. Midterm Exam 3/13/18. Thursday, March 15, 7:30 9:30 p.m. room 125 Ag Hall Midterm Exam Thursday, March 15, 7:30 9:30 p.m. room 125 Ag Hall Covers topics through Decision Trees and Random Forests (does not include constraint satisfaction) Closed book 8.5 x 11 sheet with notes

More information

Stochastic Function Norm Regularization of DNNs

Stochastic Function Norm Regularization of DNNs Stochastic Function Norm Regularization of DNNs Amal Rannen Triki Dept. of Computational Science and Engineering Yonsei University Seoul, South Korea amal.rannen@yonsei.ac.kr Matthew B. Blaschko Center

More information

Convolutional Neural Networks

Convolutional Neural Networks Lecturer: Barnabas Poczos Introduction to Machine Learning (Lecture Notes) Convolutional Neural Networks Disclaimer: These notes have not been subjected to the usual scrutiny reserved for formal publications.

More information

Data Mining. Introduction. Hamid Beigy. Sharif University of Technology. Fall 1394

Data Mining. Introduction. Hamid Beigy. Sharif University of Technology. Fall 1394 Data Mining Introduction Hamid Beigy Sharif University of Technology Fall 1394 Hamid Beigy (Sharif University of Technology) Data Mining Fall 1394 1 / 20 Table of contents 1 Introduction 2 Data mining

More information

arxiv: v1 [cs.lg] 6 Dec 2017

arxiv: v1 [cs.lg] 6 Dec 2017 This paper will appear in the proceedings of DATE 2018. Pre-print version, for personal use only. HyperPower: Power- and Memory-Constrained Hyper-Parameter Optimization for Neural Networks Dimitrios Stamoulis,

More information

Supervised Learning for Image Segmentation

Supervised Learning for Image Segmentation Supervised Learning for Image Segmentation Raphael Meier 06.10.2016 Raphael Meier MIA 2016 06.10.2016 1 / 52 References A. Ng, Machine Learning lecture, Stanford University. A. Criminisi, J. Shotton, E.

More information

Lecture 20: Neural Networks for NLP. Zubin Pahuja

Lecture 20: Neural Networks for NLP. Zubin Pahuja Lecture 20: Neural Networks for NLP Zubin Pahuja zpahuja2@illinois.edu courses.engr.illinois.edu/cs447 CS447: Natural Language Processing 1 Today s Lecture Feed-forward neural networks as classifiers simple

More information

arxiv: v2 [stat.ml] 13 Jul 2018 Abstract

arxiv: v2 [stat.ml] 13 Jul 2018 Abstract ICML 2018 AutoML Workshop Automatic Gradient Boosting Janek Thomas janek.thomas@stat.uni-muenchen.de Stefan Coors stefan.coors@campus.lmu.de Bernd Bischl bernd.bischl@stat.uni-muenchen.de Department of

More information

Progressive Neural Architecture Search

Progressive Neural Architecture Search Progressive Neural Architecture Search Chenxi Liu, Barret Zoph, Maxim Neumann, Jonathon Shlens, Wei Hua, Li-Jia Li, Li Fei-Fei, Alan Yuille, Jonathan Huang, Kevin Murphy 09/10/2018 @ECCV 1 Outline Introduction

More information

Naïve Bayes for text classification

Naïve Bayes for text classification Road Map Basic concepts Decision tree induction Evaluation of classifiers Rule induction Classification using association rules Naïve Bayesian classification Naïve Bayes for text classification Support

More information

DECISION TREES & RANDOM FORESTS X CONVOLUTIONAL NEURAL NETWORKS

DECISION TREES & RANDOM FORESTS X CONVOLUTIONAL NEURAL NETWORKS DECISION TREES & RANDOM FORESTS X CONVOLUTIONAL NEURAL NETWORKS Deep Neural Decision Forests Microsoft Research Cambridge UK, ICCV 2015 Decision Forests, Convolutional Networks and the Models in-between

More information

Classification: Linear Discriminant Functions

Classification: Linear Discriminant Functions Classification: Linear Discriminant Functions CE-725: Statistical Pattern Recognition Sharif University of Technology Spring 2013 Soleymani Outline Discriminant functions Linear Discriminant functions

More information

Modeling and Reasoning with Bayesian Networks. Adnan Darwiche University of California Los Angeles, CA

Modeling and Reasoning with Bayesian Networks. Adnan Darwiche University of California Los Angeles, CA Modeling and Reasoning with Bayesian Networks Adnan Darwiche University of California Los Angeles, CA darwiche@cs.ucla.edu June 24, 2008 Contents Preface 1 1 Introduction 1 1.1 Automated Reasoning........................

More information

The Mathematics Behind Neural Networks

The Mathematics Behind Neural Networks The Mathematics Behind Neural Networks Pattern Recognition and Machine Learning by Christopher M. Bishop Student: Shivam Agrawal Mentor: Nathaniel Monson Courtesy of xkcd.com The Black Box Training the

More information

CNNS FROM THE BASICS TO RECENT ADVANCES. Dmytro Mishkin Center for Machine Perception Czech Technical University in Prague

CNNS FROM THE BASICS TO RECENT ADVANCES. Dmytro Mishkin Center for Machine Perception Czech Technical University in Prague CNNS FROM THE BASICS TO RECENT ADVANCES Dmytro Mishkin Center for Machine Perception Czech Technical University in Prague ducha.aiki@gmail.com OUTLINE Short review of the CNN design Architecture progress

More information

ECG782: Multidimensional Digital Signal Processing

ECG782: Multidimensional Digital Signal Processing ECG782: Multidimensional Digital Signal Processing Object Recognition http://www.ee.unlv.edu/~b1morris/ecg782/ 2 Outline Knowledge Representation Statistical Pattern Recognition Neural Networks Boosting

More information

Deep Learning. Vladimir Golkov Technical University of Munich Computer Vision Group

Deep Learning. Vladimir Golkov Technical University of Munich Computer Vision Group Deep Learning Vladimir Golkov Technical University of Munich Computer Vision Group 1D Input, 1D Output target input 2 2D Input, 1D Output: Data Distribution Complexity Imagine many dimensions (data occupies

More information

Practical Methodology. Lecture slides for Chapter 11 of Deep Learning Ian Goodfellow

Practical Methodology. Lecture slides for Chapter 11 of Deep Learning  Ian Goodfellow Practical Methodology Lecture slides for Chapter 11 of Deep Learning www.deeplearningbook.org Ian Goodfellow 2016-09-26 What drives success in ML? Arcane knowledge of dozens of obscure algorithms? Mountains

More information

Bayesian Methods in Vision: MAP Estimation, MRFs, Optimization

Bayesian Methods in Vision: MAP Estimation, MRFs, Optimization Bayesian Methods in Vision: MAP Estimation, MRFs, Optimization CS 650: Computer Vision Bryan S. Morse Optimization Approaches to Vision / Image Processing Recurring theme: Cast vision problem as an optimization

More information

Escaping Local Optima: Genetic Algorithm

Escaping Local Optima: Genetic Algorithm Artificial Intelligence Escaping Local Optima: Genetic Algorithm Dae-Won Kim School of Computer Science & Engineering Chung-Ang University We re trying to escape local optima To achieve this, we have learned

More information

Neural Network Optimization and Tuning / Spring 2018 / Recitation 3

Neural Network Optimization and Tuning / Spring 2018 / Recitation 3 Neural Network Optimization and Tuning 11-785 / Spring 2018 / Recitation 3 1 Logistics You will work through a Jupyter notebook that contains sample and starter code with explanations and comments throughout.

More information