DATA.ML.100 Introduction to Pattern Recognition and Machine Learning (5 cr) – Introduction to PR and ML
DATA.ML.100 Introduction to Pattern
Recognition and Machine Learning (5 cr)
Introduction to PR and ML
Joni Kämäräinen
August 2021
Computing Sciences
Tampere University
1
History
Pattern Recognition and Machine Learning
Figure 1: King-Sun Fu
(1930-1985)
Figure 2: Marvin Minsky
(1927-2016)
2
Pattern Recognition and Machine Learning (cont.)
Figure 3: Top ML scientists (Google Scholar) 3
Machine Learning Software
Software 1.0 – Human writes the code
How would you implement function f (x) that sorts its input x (see
examples below)?
f ({8, 3, 7}) = {3, 7, 8}
f ({11, 10,−1}) = {−1, 10, 11}
f ({5, 4, 3, 3, 2, 1}) = {1, 2, 3, 3, 4, 5}
. (1)
4
Software 2.0 – Computer writes the code
How would you implement the following autonomous driving
functions f ?
(a) lane detection (b) vehicle detection
Figure 4: Software for autonomous cars. Code by Junsheng Fu who
studied at TAU for his MSc (Tech) and PhD degrees (code available in
his Github page) and now works for Zenseact, Sweden
5
Machine Learning Engineer
• Collects and annotates data
• Implements and tunes a suitable ML model
• Tests the model, collects more data if needed, re-trains and
fine-tunes the model
6
Recent results
Recent results
Natural language processing (NLP)
Text generationi (GPT-2)1
https://talktotransformer.com/
1Alec Radford et al. “Language Models are Unsupervised Multitask Learners”.
In: (2019)
7
https://talktotransformer.com/
Text understanding (BERT)2
https://www.pragnakalp.com/demos/BERT-NLP-QnA-Demo/
2Jacob Devlin et al. “BERT: Pre-training of Deep Bidirectional Transformers
for Language Understanding”. In: arXiv preprint arXiv:1810.04805 (2018)
8
https://www.pragnakalp.com/demos/BERT-NLP-QnA-Demo/
Recent results
Sound and speech
Speech synthesis and understanding3
3Y. Jia et al. “Direct speech-to-speech translation with a
sequence-to-sequence model”. In: Interspeech. 2019
9
Recent results
Image and video
Image captioning4
4A. Karpathy and L. Fei-Fei. “Deep Visual-Semantic Alignments for
Generating Image Descriptions”. In: CVPR. 2015
10
Image colorization5
5R. Zhang, P. Isola, and A.A. Efros. “Colorful Image Colorization”. In: ECCV.
2016
11
Realistic image synthesis6
6
Q. Chen and V. Koltun. “Photographic Image Synthesis with Cascaded Refinement Networks”. In: ICCV. 2017
12
NVidia GauGAN
13
Animation using GANs7
7A. Pumarola et al. “GANimation: Anatomically-aware Facial Animation from
a Single Image”. In: Proceedings of the European Conference on Computer
Vision (ECCV). 2018
14
Recent results
Robotics
Imitation learning8
8W. Yang et al. “Neural Network Controller for Autonomous Pile Loading
Revised”. In: IEEE Int. Conf. on Robotics and Automation (ICRA). Xi’an,
China, 2021. url: http://arxiv.org/abs/2103.12379
15
http://arxiv.org/abs/2103.12379
Reinforcement learning9
9A. Dag et al. “Monolithic vs. Hybrid Controller for Multi-objective
Sim-to-Real Learning”. In: IEEE/RSJ International Conference on Intelligent
Robots and Systems (IROS). Prague, Czech Rep., 2021
16
Summary
• What is the difference between Software 1.0 and 2.0?
17
History
Machine Learning Software
Recent results
Natural language processing (NLP)
Sound and speech
Image and video
Robotics