Incremental Learning Through Deep Adaptation

Given an existing trained neural network, it is often desirable to learn new capabilities without hindering performance of those already learned. Existing approaches either learn sub-optimal solutions, require joint training, or incur a substantial increment in the number of parameters for each added domain, typically as many as the original network. We propose a method called Deep Adaptation Modules (DAM) that constrains newly learned filters to be linear combinations of existing ones. DAMs precisely preserve performance on the original domain, require a fraction (typically 13 percent, dependent on network architecture) of the number of parameters compared to standard fine-tuning procedures and converge in less cycles of training to a comparable or better level of performance. When coupled with standard network quantization techniques, we further reduce the parameter cost to around 3 percent of the original with negligible or no loss in accuracy. The learned architecture can be controlled to switch between various learned representations, enabling a single network to solve a task from multiple different domains. We conduct extensive experiments showing the effectiveness of our method on a range of image classification tasks and explore different aspects of its behavior.

[1]  Geoffrey E. Hinton,et al.  ImageNet classification with deep convolutional neural networks , 2012, Commun. ACM.

[2]  Michael I. Jordan,et al.  Learning Transferable Features with Deep Adaptation Networks , 2015, ICML.

[3]  Song Han,et al.  Deep Compression: Compressing Deep Neural Network with Pruning, Trained Quantization and Huffman Coding , 2015, ICLR.

[4]  Nathan Srebro,et al.  The Marginal Value of Adaptive Gradient Methods in Machine Learning , 2017, NIPS.

[5]  Andrew Y. Ng,et al.  Reading Digits in Natural Images with Unsupervised Feature Learning , 2011 .

[6]  Kaushik Roy,et al.  Incremental Learning in Deep Convolutional Neural Networks Using Partial Network Sharing , 2017, IEEE Access.

[7]  Nikos Komodakis,et al.  Wide Residual Networks , 2016, BMVC.

[8]  Shane Legg,et al.  Human-level control through deep reinforcement learning , 2015, Nature.

[9]  Trevor Darrell,et al.  Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation , 2013, 2014 IEEE Conference on Computer Vision and Pattern Recognition.

[10]  Cheng Li,et al.  Face alignment by coarse-to-fine shape searching , 2015, 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[11]  Christoph Schroth,et al.  Car-to-Car Communication , 2006 .

[12]  Martial Hebert,et al.  Cross-Stitch Networks for Multi-task Learning , 2016, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[13]  Satish V. Ukkusuri,et al.  Modeling of Motorist-Pedestrian Interaction at Uncontrolled Mid-block Crosswalks , 2003 .

[14]  Rob Fergus,et al.  Predicting Depth, Surface Normals and Semantic Labels with a Common Multi-scale Convolutional Architecture , 2014, 2015 IEEE International Conference on Computer Vision (ICCV).

[15]  Michael S. Bernstein,et al.  ImageNet Large Scale Visual Recognition Challenge , 2014, International Journal of Computer Vision.

[16]  Derek Hoiem,et al.  Learning without Forgetting , 2016, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[17]  Robert B. Noland,et al.  Behavioural Issues in Pedestrian Speed Choice and Street Crossing Behaviour: A Review , 2008 .

[18]  Jian Sun,et al.  Deep Residual Learning for Image Recognition , 2015, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[19]  Razvan Pascanu,et al.  Progressive Neural Networks , 2016, ArXiv.

[20]  Jian Sun,et al.  Identity Mappings in Deep Residual Networks , 2016, ECCV.

[21]  Andrea Vedaldi,et al.  Integrated perception with recurrent multi-task neural networks , 2016, NIPS.

[22]  Joshua B. Tenenbaum,et al.  Human-level concept learning through probabilistic program induction , 2015, Science.

[23]  Iasonas Kokkinos,et al.  Describing Textures in the Wild , 2013, 2014 IEEE Conference on Computer Vision and Pattern Recognition.

[24]  Ioannis Mitliagkas,et al.  YellowFin and the Art of Momentum Tuning , 2017, MLSys.

[25]  Andrew Zisserman,et al.  Automated Flower Classification over a Large Number of Classes , 2008, 2008 Sixth Indian Conference on Computer Vision, Graphics & Image Processing.

[26]  Marc Alexa,et al.  How do humans sketch objects? , 2012, ACM Trans. Graph..

[27]  Subhransu Maji,et al.  Fine-Grained Visual Classification of Aircraft , 2013, ArXiv.

[28]  Xuhong Li,et al.  Explicit Inductive Bias for Transfer Learning with Convolutional Networks , 2018, ICML.

[29]  Mubarak Shah,et al.  UCF101: A Dataset of 101 Human Actions Classes From Videos in The Wild , 2012, ArXiv.

[30]  Leonidas J. Guibas,et al.  Taskonomy: Disentangling Task Transfer Learning , 2018, 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.

[31]  Andrea Vedaldi,et al.  Universal representations: The missing link between faces, text, planktons, and cat breeds , 2017, ArXiv.

[32]  Trevor Darrell,et al.  Fully Convolutional Networks for Semantic Segmentation , 2017, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[33]  Dariu Gavrila,et al.  Context-Based Pedestrian Path Prediction , 2014, ECCV.

[34]  Trevor Darrell,et al.  DeCAF: A Deep Convolutional Activation Feature for Generic Visual Recognition , 2013, ICML.

[35]  Erich Elsen,et al.  Deep Speech: Scaling up end-to-end speech recognition , 2014, ArXiv.

[36]  Dariu Gavrila,et al.  Analysis of pedestrian dynamics from a vehicle perspective , 2014, 2014 IEEE Intelligent Vehicles Symposium Proceedings.

[37]  Andrea Vedaldi,et al.  Learning multiple visual domains with residual adapters , 2017, NIPS.

[38]  Feng Jiang,et al.  Pedestrian behavior analysis using 110-car naturalistic driving data in USA , 2013 .

[39]  Mike McDonald,et al.  Study of pedestrians' gap acceptance behavior when they jaywalk outside crossing facilities , 2010, 13th International IEEE Conference on Intelligent Transportation Systems.

[40]  Patrick Heinemann,et al.  Context-based detection of pedestrian crossing intention for autonomous driving in urban environments , 2016, 2016 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS).

[41]  George Yannis,et al.  A critical assessment of pedestrian behaviour models , 2009 .

[42]  Razvan Pascanu,et al.  Overcoming catastrophic forgetting in neural networks , 2016, Proceedings of the National Academy of Sciences.

[43]  Anders Lindgren,et al.  Requirements for the Design of Advanced Driver Assistance Systems - The Differences between Swedish and Chinese Drivers , 2008 .

[44]  Ross B. Girshick,et al.  Mask R-CNN , 2017, 1703.06870.

[45]  Mohan M. Trivedi,et al.  Trajectory analysis and prediction for improved pedestrian safety: Integrated framework and evaluations , 2015, 2015 IEEE Intelligent Vehicles Symposium (IV).

[46]  Stefan Carlsson,et al.  CNN Features Off-the-Shelf: An Astounding Baseline for Recognition , 2014, 2014 IEEE Conference on Computer Vision and Pattern Recognition Workshops.

[47]  Dariu Gavrila,et al.  An Experimental Study on Pedestrian Classification , 2006, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[48]  Andrew Zisserman,et al.  Very Deep Convolutional Networks for Large-Scale Image Recognition , 2014, ICLR.

[49]  Joan Bruna,et al.  Exploiting Linear Structure Within Convolutional Networks for Efficient Evaluation , 2014, NIPS.

[50]  G. Griffin,et al.  Caltech-256 Object Category Dataset , 2007 .

[51]  Trevor Darrell,et al.  Adversarial Discriminative Domain Adaptation , 2017, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[52]  R. French Catastrophic forgetting in connectionist networks , 1999, Trends in Cognitive Sciences.

[53]  D. Clay,et al.  Driver attitude and attribution : implications for accident prevention , 1995 .

[54]  Lorien Y. Pratt,et al.  Comparing Biases for Minimal Network Construction with Back-Propagation , 1988, NIPS.

[55]  Bernhard Schölkopf,et al.  A Kernel Two-Sample Test , 2012, J. Mach. Learn. Res..

[56]  Jimmy Ba,et al.  Adam: A Method for Stochastic Optimization , 2014, ICLR.

[57]  Alex Krizhevsky,et al.  Learning Multiple Layers of Features from Tiny Images , 2009 .

[58]  David Hsu,et al.  Intention-aware online POMDP planning for autonomous driving in a crowd , 2015, 2015 IEEE International Conference on Robotics and Automation (ICRA).

[59]  Ralf Risser,et al.  Pedestrian-driver communication and decision strategies at marked crossings. , 2017, Accident; analysis and prevention.

[60]  Li-Ta Hsu,et al.  Probability estimation for pedestrian crossing intention at signalized crosswalks , 2015, 2015 IEEE International Conference on Vehicular Electronics and Safety (ICVES).

[61]  Bram van Ginneken,et al.  A survey on deep learning in medical image analysis , 2017, Medical Image Anal..

[62]  Roland Siegwart,et al.  A data-driven approach for pedestrian intention estimation , 2016, 2016 IEEE 19th International Conference on Intelligent Transportation Systems (ITSC).

[63]  Johannes Stallkamp,et al.  Man vs. computer: Benchmarking machine learning algorithms for traffic sign recognition , 2012, Neural Networks.

[64]  Michael I. Jordan,et al.  Unsupervised Domain Adaptation with Residual Transfer Networks , 2016, NIPS.

[65]  Song Han,et al.  Learning both Weights and Connections for Efficient Neural Network , 2015, NIPS.

[66]  C. Michaels,et al.  To Cross or Not to Cross: The Effect of Locomotion on Street-Crossing Behavior , 1996 .

[67]  Christoph H. Lampert,et al.  Learning to detect unseen object classes by between-class attribute transfer , 2009, 2009 IEEE Conference on Computer Vision and Pattern Recognition.