Learning on Humanoid Robots Vadym Gryshchuk 19.11.2018 Outline - - PowerPoint PPT Presentation

▶
learning on humanoid robots
SMART_READER_LITE
LIVE PREVIEW

Learning on Humanoid Robots Vadym Gryshchuk 19.11.2018 Outline - - PowerPoint PPT Presentation

Sep 20, 2022 246 likes •585 views

Neural Architectures for Lif ifelong Learning on Humanoid Robots Vadym Gryshchuk 19.11.2018 Outline Motivation Background Approaches Results Discussion Conclusion Neural Architectures for Lifelong Learning on Humanoid

slide-1
SLIDE 1

Neural Architectures for Lif ifelong Learning on Humanoid Robots

Vadym Gryshchuk 19.11.2018

slide-2
SLIDE 2

Outline

  • Motivation
  • Background
  • Approaches
  • Results
  • Discussion
  • Conclusion

2 Neural Architectures for Lifelong Learning on Humanoid Robots

slide-3
SLIDE 3
  • Motivation
  • Background
  • Approaches
  • Results
  • Discussion
  • Conclusion

3 Neural Architectures for Lifelong Learning on Humanoid Robots

slide-4
SLIDE 4

What is is Lif ifelong Learning?

  • Continual acquisition of knowledge
  • Fine-tuning of knowledge
  • Learning from experiences
  • Retaining of previously learnt

experiences

Neural Architectures for Lifelong Learning on Humanoid Robots 4

Figure 1.1: NICO – Neuro-Inspired COmpanion (Source: Kerzel et al. [2]).

slide-5
SLIDE 5

Catastrophic Forgetting

  • Interference of learnt representations with new information

Neural Architectures for Lifelong Learning on Humanoid Robots 5

Representation 1 Representation 2

slide-6
SLIDE 6

In Inspiration from Biological Systems

  • Neurosynaptic plasticity
  • Hippocampus and cerebral cortex
  • Transfer learning
  • Intrinsic motivation
  • Crossmodal learning
  • Incremental learning

Neural Architectures for Lifelong Learning on Humanoid Robots 6

slide-7
SLIDE 7
  • Motivation
  • Background
  • Approaches
  • Results
  • Discussion
  • Conclusion

7 Neural Architectures for Lifelong Learning on Humanoid Robots

slide-8
SLIDE 8

Neural Networks

Neural Architectures for Lifelong Learning on Humanoid Robots 8

Figure 2.1: Neural network representation (Source: McDonald [3]).

slide-9
SLIDE 9

Convolutional Neural l Networks (C (CNNs)

Neural Architectures for Lifelong Learning on Humanoid Robots 9

Figure 2.2: Convolutional neural network (Source: Cavaioni [1])

slide-10
SLIDE 10

Self lf-Organizing Networks

  • Self-Organizing Map (SOM)
  • Grow When Required Network (GWR Network)
  • Recurrent GWR

Neural Architectures for Lifelong Learning on Humanoid Robots 10

slide-11
SLIDE 11
  • Motivation
  • Background
  • Approaches
  • Results
  • Discussion
  • Conclusion

11 Neural Architectures for Lifelong Learning on Humanoid Robots

slide-12
SLIDE 12

Object Recognition: CNN + Cla lassifier

  • Learning from video sequences
  • Visual transformations of objects
  • Changing environment

Neural Architectures for Lifelong Learning on Humanoid Robots 12

Figure 3.1: iCub (Source: Pasquale et al. [6]).

slide-13
SLIDE 13

Object Recognition: CNN + Classifier

Neural Architectures for Lifelong Learning on Humanoid Robots 13

CNN

Classifier

apple cup ball tomato

slide-14
SLIDE 14

iC iCub: Object Learning

Neural Architectures for Lifelong Learning on Humanoid Robots 14

Source: https://www.youtube.com/watch?v=ghUFweqm7W8

slide-15
SLIDE 15

iCub: Object Learning

Neural Architectures for Lifelong Learning on Humanoid Robots 15

Source: https://www.youtube.com/watch?v=ghUFweqm7W8

slide-16
SLIDE 16

iCub: Object Learning

Neural Architectures for Lifelong Learning on Humanoid Robots 16

Source: https://www.youtube.com/watch?v=ghUFweqm7W8

slide-17
SLIDE 17

iCub: Object Learning

Neural Architectures for Lifelong Learning on Humanoid Robots 17

Source: https://www.youtube.com/watch?v=ghUFweqm7W8

slide-18
SLIDE 18

iCub: Object Learning

Neural Architectures for Lifelong Learning on Humanoid Robots 18

Source: https://www.youtube.com/watch?v=ghUFweqm7W8

slide-19
SLIDE 19

iCub: Object Learning

Neural Architectures for Lifelong Learning on Humanoid Robots 19

Source: https://www.youtube.com/watch?v=ghUFweqm7W8

slide-20
SLIDE 20

Sensorimotor Learning: Self lf-Organization

  • Latency in sensorimotor systems
  • Predictive mechanisms for future

motor states

  • Online learning

Neural Architectures for Lifelong Learning on Humanoid Robots 20

Source: https://upload.wikimedia.org/wikipedia/commons/ 4/47/Nao_Robot_%28Robocup_2016%29.jpg

slide-21
SLIDE 21

Sensorimotor Learning: Self-Organization

Neural Architectures for Lifelong Learning on Humanoid Robots 21

Figure 3.2: The imitation scenario (Source: Mici et al. [4]).

slide-22
SLIDE 22

Sensorimotor Learning: Self-Organization

Neural Architectures for Lifelong Learning on Humanoid Robots 22

Figure 3.3: Visuomotor learning (Source: Mici et al. [4]).

slide-23
SLIDE 23

Object Recognition: CNN + Self-Organization

Neural Architectures for Lifelong Learning on Humanoid Robots 23

RGB sequence Pre-trained CNN Pre-trained CNN Depth sequence Features Self-organizing network Label

Figure 3.4: Recognition pipeline (Adapted from Part et al. [5]).

slide-24
SLIDE 24
  • Motivation
  • Background
  • Approaches
  • Results
  • Discussion
  • Conclusion

24 Neural Architectures for Lifelong Learning on Humanoid Robots

slide-25
SLIDE 25

Object Recognition: CNN + Cla lassifier

Neural Architectures for Lifelong Learning on Humanoid Robots 25

Figure 4.1: Classification accuracy of the model, which was trained on an incremental number of objects (Source: Pasquale et al. [6]).

slide-26
SLIDE 26

Object Recognition: CNN + Cla lassifier

Neural Architectures for Lifelong Learning on Humanoid Robots 26

Figure 4.2: Classification accuracy of the model trained incrementally on different days (Source: Pasquale et al. [6]).

slide-27
SLIDE 27

Sensorimotor Learning: : Self-Organizing Architecture

Neural Architectures for Lifelong Learning on Humanoid Robots 27

Figure 4.3: Behaviour of the architecture (Source: Mici et al. [4]).

slide-28
SLIDE 28

Object Recognition: CNN + Self-Organization

Neural Architectures for Lifelong Learning on Humanoid Robots 28

Figure 4.4: Recognition pipeline (Source: Part et al. [5]).

slide-29
SLIDE 29

Dis iscussion

  • CNN + Classifier architecture for object recognition:
  • Features extracted from a CNN are dependent on a dataset the model was

trained on

  • Old representations are overwritten by the new information
  • Self-organizing architecture for sensorimotor learning:
  • Incremental online learning and prediction
  • Unreliability of visual body tracking framework in complex body positions
  • CNN + self-organization for object recognition:
  • Self-organizing network grows when required
  • Temporal context is not considered

Neural Architectures for Lifelong Learning on Humanoid Robots 29

slide-30
SLIDE 30

Conclusion

  • Lifelong learning is crucial for intelligent robots
  • Biological systems provide a basis for the incremental learning
  • Self-organizing networks preserve the topology
  • CNNs learn efficient feature descriptors
  • Catastrophic forgetting increases during incremental tasks

Neural Architectures for Lifelong Learning on Humanoid Robots 30

slide-31
SLIDE 31

Thank You! Questions?

Neural Architectures for Lifelong Learning on Humanoid Robots 31

slide-32
SLIDE 32

References

  • [1] Cavaioni, M. Deep Learning series: Convolutional Neural Networks. https://medium.com/machine- learning-

bites/deeplearning-series-convolutional-neural-networks-a9c2f2ee1524 . [Online; accessed 13-November-2018].

  • [2] Kerzel, M., Strahl, E., Magg, S., Navarro-Guerrero, N., Heinrich, S., Wermter, S.NICO - neuro-inspired

companion: A developmental humanoid robot platform for multimodal interaction. In26th IEEE International Symposium on Robot and Human Interactive Communication, ROMAN 2017, Lisbon, Portugal, August 28 - Sept. 1, 2017, pages 113–120, 2017.

  • [3] McDonald, C. Machine learning fundamentals (II): Neural networks. https://towardsdatascience.com/machine-

learning-fundamentals-ii-neural-networks-f1e7b2cb3eef. [Online; accessed 13-November-2018].

  • [4] Mici, L., Parisi, I. G., Wermter, S. An Incremental Self-Organizing Architecture for Sensorimotor Learning and
  • Prediction. CoRR, abs/1712.08521, 2017.
  • [5] Part l. J., Lemon, O. Incremental online learning of objects for robots operating in real environments. In Joint

IEEE International Conference on Development and Learning and Epigenetic Robotics, ICDL-EpiRob 2017, Lisbon, Portugal, September 18-21, 2017, pages 304–310, 2017.

  • [6] Pasquale, G., Ciliberto, C., Odone, F., Rosasco, L., Natale, L. Real-world object recognition with off-the-shelf deep

conv nets: How many objects can iCub learn? CoRR, abs/1504.03154, 2015.

Neural Architectures for Lifelong Learning on Humanoid Robots 32