learning with bad training data via iterative trimmed
play

Learning with Bad Training Data via Iterative Trimmed Loss - PowerPoint PPT Presentation

Mar 25, 2024 •435 likes •532 views

Learning with Bad Training Data via Iterative Trimmed Loss Minimization Yanyao Shen , Sujay Sanghavi University of Texas at Austin Motivations 1: Bad Training Labels in Classification Supervised: noise in training labels makes classifiers

  1. Learning with Bad Training Data via Iterative Trimmed Loss Minimization Yanyao Shen , Sujay Sanghavi University of Texas at Austin

  2. Motivations 1: Bad Training Labels in Classification Supervised: noise in training labels makes classifiers inaccurate 2. bird Systematic label noise: a fraction of “horse” is 7. horse mis-labeled “bird” Dataset size will not 2. bird rescue …

  3. Motivations 1: Bad Training Labels in Classification 2: Mixed Training Data Supervised: noise in training labels Unsupervised: spurious samples give makes classifiers inaccurate bad generative models 2. bird Systematic label noise: a fraction of “horse” is 7. horse mis-labeled “bird” Dataset size will not 2. bird rescue … + GAN =

  4. Motivations 1: Bad Training Labels in Classification 2: Mixed Training Data Supervised: noise in training labels Unsupervised: spurious samples give makes classifiers inaccurate bad generative models 2. bird Systematic label noise: a fraction of “horse” is 7. horse mis-labeled “bird” Dataset size will not 2. bird rescue … + GAN = 3: Backdoor Attacks Images classified as `ship’ Images classified as `horse’

  5. Observation: Initial Epochs Can Differentiate

Recommend

PERFORMANCE S&Bs Community Bank benchmark includes banks with assets between $100 million

PERFORMANCE S&Bs Community Bank benchmark includes banks with assets between $100 million

2018 Q4 COMMUNITY BANK PERFORMANCE S&Bs Community Bank benchmark includes banks with assets between $100 million and $5 billion. 69.05% 69.37% 5% Trimmed Mean 5% Trimmed Mean 5% Trimmed Mean 5% Trimmed Mean 1.15% 5% Trimmed Mean

202 views • 19 slides
www.mpc.berkeley.edu Borrelli (UC Berkeley) Iterative Learning MPC 2018 CDC Slide 1

www.mpc.berkeley.edu Borrelli (UC Berkeley) Iterative Learning MPC 2018 CDC Slide 1

Iterative Learning Model Predictive Control Francesco Borrelli Email: fborrelli@berkeley.edu University of California Berkeley, USA www.mpc.berkeley.edu Borrelli (UC Berkeley) Iterative Learning MPC 2018 CDC Slide 1 Acknowledgements

1.19k views • 104 slides
Iterative Learning of Feed Forward Corrections for High Performance Tracking Fabian L.

Iterative Learning of Feed Forward Corrections for High Performance Tracking Fabian L.

Iterative Learning of Feed Forward Corrections for High Performance Tracking Fabian L. Mueller, Angela P. Schoellig, Raffaello DAndrea Institute for Dynamic Systems and Control ETH Zrich, Switzerland 1 Iterative Learning of Feed

500 views • 15 slides
Edge-based Discovery of Training Data for Machine Learning CMU authors Deep Learning Recipe

Edge-based Discovery of Training Data for Machine Learning CMU authors Deep Learning Recipe

CSci 8980 Edge-based Discovery of Training Data for Machine Learning CMU authors Deep Learning Recipe Collect a large amount of data and label it Select a model and train a DNN Deploy the DNN for inference Labelled Data Some

409 views • 14 slides
Learning From Data Lecture 5 Training Versus Testing The Two Questions of Learning Theory of

Learning From Data Lecture 5 Training Versus Testing The Two Questions of Learning Theory of

Learning From Data Lecture 5 Training Versus Testing The Two Questions of Learning Theory of Generalization ( E in E out ) An Effective Number of Hypotheses A Combinatorial Puzzle M. Magdon-Ismail CSCI 4100/6100 recap: The Two Questions

700 views • 18 slides
Automatic Scaling Iterative Computations Guozhang Wang Cornell University Aug. 7 th , 2012

Automatic Scaling Iterative Computations Guozhang Wang Cornell University Aug. 7 th , 2012

Automatic Scaling Iterative Computations Guozhang Wang Cornell University Aug. 7 th , 2012 1 What are Non-Iterative Computations? Non-iterative computation flow Input Data Directed Acyclic Operator 1 Examples Batch style

664 views • 44 slides
Re Revie view w of Event nt Based d Trimmed med Meshing ing for In-Cy Cylind linder er

Re Revie view w of Event nt Based d Trimmed med Meshing ing for In-Cy Cylind linder er

Re Revie view w of Event nt Based d Trimmed med Meshing ing for In-Cy Cylind linder er Simula ulation tion with STAR AR-CCM CCM+ Gerald ld Schmi hmidt, t, CD-adapco dapco In In-Cylind Cylinder er Develop elopment ment Pro

330 views • 8 slides
Iterative Ensemble Classification for Relational Data A Case Study of Semantic Web Services

Iterative Ensemble Classification for Relational Data A Case Study of Semantic Web Services

Outline Introduction Iterative Ensemble Classification Conclusion Iterative Ensemble Classification for Relational Data A Case Study of Semantic Web Services Andreas He and Nicholas Kushmerick University College Dublin, Ireland September

838 views • 51 slides
Co-Training Based on Combining Labeled and Unlabeled Data with Co-Training by A. Blum

Co-Training Based on Combining Labeled and Unlabeled Data with Co-Training by A. Blum

0. Co-Training Based on Combining Labeled and Unlabeled Data with Co-Training by A. Blum & T. Mitchell, 1998 1. Problem: Learning to classify data (ex: web pages) when the description of each example can be partitioned in 2

560 views • 9 slides
Coresets for Data-efficient Training of Machine Learning Models Baharan Mirzasoleiman*, Jeff

Coresets for Data-efficient Training of Machine Learning Models Baharan Mirzasoleiman*, Jeff

Coresets for Data-efficient Training of Machine Learning Models Baharan Mirzasoleiman*, Jeff Bilmes**, Jure Leskovec* * * ** Machine Learning Becomes Mainstream Personalized medicine Robotics Finance Autonomous cars 2 Data is the Fuel

704 views • 22 slides
MetaFun: Meta-Learning with Iterative Functional Updates Jin Xu, Jean-Francois Ton, Hyunjik Kim,

MetaFun: Meta-Learning with Iterative Functional Updates Jin Xu, Jean-Francois Ton, Hyunjik Kim,

MetaFun: Meta-Learning with Iterative Functional Updates Jin Xu, Jean-Francois Ton, Hyunjik Kim, Adam R. Kosiorek, Yee Whye Teh 37th International Conference on Machine Learning Supervised Meta-Learning Supervised Meta-Learning Supervised

1.01k views • 66 slides
Learning Training data consists of pairs of { X n , Y n } w T w T S w ( X, Y ) = y i x i

Learning Training data consists of pairs of { X n , Y n } w T w T S w ( X, Y ) = y i x i

Learning Training data consists of pairs of { X n , Y n } w T w T S w ( X, Y ) = y i x i + y i ,y j d ij i i,j S w ( X, Y ) = w T ( X, Y ) 16 Learning with SVMs 1 2 w T w argmin w w T ( X n , Y n ) w T ( X n , H n )

439 views • 9 slides
Learning Normalized Inputs for Iterative Estimation in Medical Image Segmentation Michal

Learning Normalized Inputs for Iterative Estimation in Medical Image Segmentation Michal

4/6/18 Learning Normalized Inputs for Iterative Estimation in Medical Image Segmentation Michal Drozdzal, Gabriel Chartrand, Eugene Vorontsov, Mahsa Shakeri, Lisa Di Jorio, An Tang, Adriana Romero, Yoshua Bengio, Chris Pal, Samuel Kadoury

165 views • 12 slides
S8941 SYNTHETIC LABEL DATA FOR TRAINING DEEP LEARNING ISR ALGORITHMS WILL RORRER, PROGRAM

S8941 SYNTHETIC LABEL DATA FOR TRAINING DEEP LEARNING ISR ALGORITHMS WILL RORRER, PROGRAM

Place image here (13.33 x 3.5) S8941 SYNTHETIC LABEL DATA FOR TRAINING DEEP LEARNING ISR ALGORITHMS WILL RORRER, PROGRAM MANAGER NVIDIA GTC San Jose 29 March 2018 NON-EXPORT CONTROLLED THESE ITEM(S) / DATA HAVE BEEN REVIEWED IN

551 views • 51 slides
Which data do we need for training? Domain Adaption and Learning under Label Noise Franz

Which data do we need for training? Domain Adaption and Learning under Label Noise Franz

56 th Photogrammetric Week Stuttgart 20170913 Which data do we need for training? Domain Adaption and Learning under Label Noise Franz Rottensteiner Institute of Photogrammetry and GeoInformation Leibniz Universitt Hannover

1.21k views • 53 slides
social data science Statistical Learning Sebastian Barfort August 15, 2016 University of

social data science Statistical Learning Sebastian Barfort August 15, 2016 University of

social data science Statistical Learning Sebastian Barfort August 15, 2016 University of Copenhagen Department of Economics 1/85 concepts Cross validation: Split data in test and training data. Train model on training data, test it on test

884 views • 85 slides
Introduction to Deep Learning 1 / 24 Is it a question? Given training data with categories A (

Introduction to Deep Learning 1 / 24 Is it a question? Given training data with categories A (

Introduction to Deep Learning 1 / 24 Is it a question? Given training data with categories A ( ) and B ( ), say well drilling sites with different outcomes 2 / 24 Is it a question? Given training data with categories A ( ) and B (

866 views • 49 slides
Design of Norm-Optimal Iterative Learning Controllers: The Effect of an Iteration-Domain Kalman

Design of Norm-Optimal Iterative Learning Controllers: The Effect of an Iteration-Domain Kalman

Design of Norm-Optimal Iterative Learning Controllers: The Effect of an Iteration-Domain Kalman Filter for Disturbance Estimation Nicolas Degen, Autonomous System Lab, ETH Zrich Angela P. Schoellig, University of Toronto Institute of Aerospace

379 views • 15 slides
Iterative learning control (Study of work by Christian Schmidt and others) M.Musienko, USPAS

Iterative learning control (Study of work by Christian Schmidt and others) M.Musienko, USPAS

Iterative learning control (Study of work by Christian Schmidt and others) M.Musienko, USPAS 2017 Free Electron Laser in Hamburg (FLASH) at DESY pulsed RF Operation due to the thermal losses FLASH LLRF Disturbances - microphonic

331 views • 22 slides
Grid-Based SAT Solving with Iterative Partitioning and Clause Learning Antti E. J. Hyv arinen,

Grid-Based SAT Solving with Iterative Partitioning and Clause Learning Antti E. J. Hyv arinen,

Grid-Based SAT Solving with Iterative Partitioning and Clause Learning Antti E. J. Hyv arinen, Tommi Junttila, and Ilkka Niemel a Department of Information and Computer Science Aalto University, School of Science Finland September 16,

453 views • 17 slides
Optimization Based Iterative Learning Control for Trajectory Tracking Angela Schoellig and

Optimization Based Iterative Learning Control for Trajectory Tracking Angela Schoellig and

Optimization Based Iterative Learning Control for Trajectory Tracking Angela Schoellig and Raffaello DAndrea Institute for Dynamic Systems and Control ETH Zrich, Switzerland European Control Conference 2009 Budapest, Hungary 1

366 views • 15 slides
Machine Learning Pipelines Marco Serafini COMPSCI 532 Lecture 21 Training vs. Inference

Machine Learning Pipelines Marco Serafini COMPSCI 532 Lecture 21 Training vs. Inference

Machine Learning Pipelines Marco Serafini COMPSCI 532 Lecture 21 Training vs. Inference Training: data model Computationally expensive No hard real-time requirements (typically) Inference: data + model prediction

432 views • 25 slides
S8906: FAST DATA PIPELINES FOR DEEP LEARNING TRAINING Przemek Tredak, Simon Layton THE PROBLEM

S8906: FAST DATA PIPELINES FOR DEEP LEARNING TRAINING Przemek Tredak, Simon Layton THE PROBLEM

S8906: FAST DATA PIPELINES FOR DEEP LEARNING TRAINING Przemek Tredak, Simon Layton THE PROBLEM 2 CPU BOTTLENECK OF DL TRAINING CPU : GPU ratio Multi-GPU, dense systems are more common (DGX-1V, DGX-2) Using more cores / sockets is very

595 views • 41 slides
Iterative Techniques in Matrix Algebra Jacobi & Gauss-Seidel Iterative Techniques II

Iterative Techniques in Matrix Algebra Jacobi & Gauss-Seidel Iterative Techniques II

Iterative Techniques in Matrix Algebra Jacobi & Gauss-Seidel Iterative Techniques II Numerical Analysis (9th Edition) R L Burden & J D Faires Beamer Presentation Slides prepared by John Carroll Dublin City University 2011

1.51k views • 107 slides

More recommend