Machine Learning and Robotics PhD positions (with scholarships) are
available at the iCub Facility and the Laboratory for Computational and
Statistical Learning (LCSL, IIT@MIT), Istituto Italiano di Tecnologia
(IIT). See list of topics below.
Check the official
https://www.iit.it/phd-school. and the tips+tricks section for the detailed instructions on how to apply.
Applications must be filed through the University of Genoa using the online service at this link:
https://www.studenti.unige.it/postlaurea/dottorati/XXXII/ENG/. Application deadline: June 10, 2016 at 12.00 noon (Italian time).
——————————
———————–
2) Scene analysis using deep-learning
Tutor: Lorenzo Natale, Lorenzo Rosasco
Description: machine learning, and in particular deep learning
methods, have been applied with remarkable success to solve visual
problems like pedestrian detection, object retrieval, recognition and
segmentation. One of the difficulties with these techniques is that
training requires a large amount of data and it is not straightforward
to adopt them when training samples are acquired online and autonomously
by a robot. One solution is to adopt pre-trained convolutional neural
networks (DCNN) for image representation and use simpler classifiers,
either in batch or incrementally. Following this approach DCNNs have
been integrated in the iCub visual system leading to a remarkable
increase of object recognition performance. However, scene analysis in
realistic settings is still challenging due to scale, light variability
and clutter. The goal of this project is to further investigate and
improve the iCub recognition and visual segmentation capabilities. To
this aim we will investigate techniques for pixel-base semantic
segmentation using DCNNs and object detection mixing top-down and
bottom-up cues for image segmentation.
Requirements: This PhD project will be carried out within the
Humanoid Sensing and Perception lab (iCub Facility) and Laboratory for
Computational and Statistical Learning. The ideal candidate should have a
degree in Computer Science or Engineering (or equivalent) and
background in Machine Learning, Robotics and possibly in Computer
Vision. He should also be highly motivated to work on a robotic platform
and have strong computer programming skills.
References:
Olga Russakovsky, Jia Deng, Hao Su, Jonathan Krause, Sanjeev
Satheesh, Sean Ma, Zhiheng Huang, Andrej Karpathy, Aditya Khosla,
Michael Bernstein, Alexander C. Berg and Li Fei-Fei, ImageNet Large
Scale Visual Recognition Challenge, arXiv:1409.0575, 2014
Jon Long, Evan Shelhamer, Trevor Darrell, Fully Convolutional Networks for Semantic Segmentation, CVPR 2015
Pasquale, G., Ciliberto, C., Odone, F., Rosasco, L., and Natale, L.,
Teaching iCub to recognize objects using deep Convolutional Neural
Networks, in Proc. 4th Workshop on Machine Learning for Interactive
Systems, 2015
3) Implicit learning
Tutor: Lorenzo Natale, Lorenzo Rosasco
Description: machine learning, and in particular deep learning
methods, have been applied with remarkable success to solve visual
problems like pedestrian detection, object retrieval, recognition and
segmentation. One of the difficulties with these techniques it that
training requires a large amount of labelled data and it is not
straightforward to adopt them when training samples are acquired online
and autonomously by the robot. Critical issues are how to obtain large
amount of training samples, how to perform object segmentation and
labelling. They key idea is develop weakly supervised frameworks, where
learning can exploit forms of implicit
labelling. In previous work we have proposed to exploit coherence
between perceived motion and the robot own-motion to autonomously learn a
visual detector of the hand. The goal of this project is to investigate
algorithms for learning to recognize object by exploiting implicit
supervision, focusing in particular on the strategies that allow the
robot to extract training samples autonomously, starting from motion and
disparity cues.
Requirements: This PhD project will be carried out within the
Humanoid Sensing and Perception lab (iCub Facility) and Laboratory for
Computational and Statistical Learning. The ideal candidate should have a
degree in Computer Science or Engineering (or equivalent) and
background in Machine Learning, Robotics and possibly in Computer
Vision. He should also be highly motivated to work on a robotic platform
and have strong computer programming skills.
References:
Olga Russakovsky, Jia Deng, Hao Su, Jonathan Krause, Sanjeev
Satheesh, Sean Ma, Zhiheng Huang, Andrej Karpathy, Aditya Khosla,
Michael Bernstein, Alexander C. Berg and Li Fei-Fei, ImageNet Large
Scale Visual Recognition Challenge, arXiv:1409.0575, 2014.
Wang, X., Gupta, A., Unsupervised Learning of Visual Representations using Videos, arXiv:1505.00687v2, 2015.
Ciliberto, C., Smeraldi, F., Natale, L., Metta, G., Online Multiple
Instance Learning Applied to Hand Detection in a Humanoid Robot,
IEEE/RSJ International Conference on Intelligent Robots and Systems, San
Francisco, California, September 25-30, 2011.
4) Learning to recognize objects using multimodal cues
Tutor: Lorenzo Natale, Lorenzo Rosasco
Description: robots can actively sense the environment using not only
vision but also haptic information. One of the problems to be addressed
in this case is how to control the robot to explore the environment and
extract relevant information (so called exploratory procedures).
Conventionally, learning to recognize objects has been primarily
addressed using vision. However, physical properties of objects are more
directly perceived using other sensory modalities. For this reason,
recent work has started to investigate how to discriminate objects using
other sensory channels, like touch, force and proprioception. The goals
of this project are i) to implement control strategies for object
exploration, investigating to what extent different
explorative strategies contribute to object discrimination, ii) the
implementation of learning algorithms that allow the robot to
discriminate objects using haptic information and, finally, iii) to
investigate how haptic information can be integrated with vision to
build a rich model of the objects for better discrimination.
Requirements: This PhD project will be carried out within the
Humanoid Sensing and Perception lab (iCub Facility) and Laboratory for
Computational and Statistical Learning. The ideal candidate should have a
degree in Computer Science or Engineering (or equivalent) and
background in Machine Learning, Robotics and possibly in Computer
Vision. He should also be highly motivated to work on a robotic platform
and have strong computer programming skills.
References:
Pasquale, G., Ciliberto, C., Odone, F., Rosasco, L., and Natale, L.,
Teaching iCub to recognize objects using deep Convolutional Neural
Networks, in Proc. 4th Workshop on Machine Learning for Interactive
Systems, 2015.
Liarokapis, M.V., Çalli, B., Spiers, A.J, Dollar, A.M., Unplanned,
model-free, single grasp object classification with underactuated hands
and force sensors, IROS, 2015.
Madry, M., Bo, L., Kragic, D. and Fox, D., ST-HMP: Unsupervised Spatio-Temporal feature learning for tactile data, ICRA 2014.
12) Tera-Scale Machine Learning in a Dynamic World
Tutors: Giorgio Metta, Lorenzo Rosasco
Department: iCub Facility
Description: Machine learning methods have been applied with
remarkable success to solve a variety of perceptual/cognitive tasks,
e.g. in vision and speech. Yet, current state of the art algorithms are
challenged by increasing amount of high-dimensional data, possibly
evolving over time. The goal of this project is to tackle the design and
deployment of sound machine learning methods to learn from large/huge
scale data sets possibly acquired over time in changing conditions. This
problem finds natural application in the context of humanoid robotics
where data are constantly made available by sensory systems. The plan is
to first consider extensions of classical convex approaches such as
kernel methods, by incorporating ideas such as sketching, hashing and
randomization. The idea is then to further explore possibly non-convex,
hierarchical (deep) learning methods, based on greedy optimization
strategy. Emphasis will be on developing effective solutions, while
keeping optimal theoretical guarantees.
Requirements: This PhD project will be carried out within the iCub
Facility and Laboratory for Computational and Statistical Learning. The
ideal candidate should have a degree in Computer Science or Engineering
(or equivalent) and background in Machine Learning, Robotics and in
optimization/signal processing. He should also be highly motivated to
work on a robotic platform and have strong mathematical and or/ computer
programming skills.
References:
Incremental Semiparametric Inverse Dynamics Learning, Raffaello
Camoriano, Silvio Traversaro, Lorenzo Rosasco, Giorgio Metta, Francesco
Nori (IROS 2016)
Less is More: Nyström Computational Regularization Alessandro Rudi, Raffaello Camoriano, Lorenzo Rosasco (NIPS 2015)
Contacts: Giorgio Metta (
giorgio[dot]metta[at]iit.it) and Lorenzo Rosasco (
lrosasco[at]mit.edu)
