Machine Learning: A gentle Introduction
Download as PPTX, PDF5 likes1,060 views
The document provides a comprehensive overview of machine learning, its concepts, and applications, including reinforcement learning techniques like Deep Q-Learning. It discusses the significance of data acquisition for successful machine learning projects and highlights various model architectures and learning types such as supervised and unsupervised learning. Additionally, it reviews the advancement of machine learning performance across multiple domains, illustrating the potential of ML to exceed human capabilities.
Machine Learning: A gentle Introduction
- 1. Machine Learning «A Gentle Introduction» @ZimMatthias Matthias Zimmermann BSI Business Systems Integration AG
- 2. «What is the difference to IBM’s chess system 20 years ago?»
- 3. AlphaGo Hardware Powered by TPUs
- 4. Tensor Processing Unit (TPU) Specialized ML Hardware What else is needed?
- 6. Markov Decision Process Environment (Atari Breakout) Agent performing Actions (Left, Right, Release Ball) State (Bricks, location / direction of ball, …) Rewards (A Brick is hit) Deep Reinforcement Learning
- 7. Q-Learning (simplified) Markov Decision Process Q(s, a) Highest sum of future Rewards for action a in state s initialize Q randomly set initial state s0 repeat execute a to maximize Q(si, a) observe r and new state si+1 set Q = update(Q, r, a, si+1) set si = si+1 until terminated Deep Reinforcement Learning
- 8. Deep Q Learning (DQN) Q Learning Q(s, a) = Deep Neural Network (DNN) Retrain DNN regularly (using it’s own experience) Deep Reinforcement Learning Action a Left, Right, Release DNN Q(s, a) State s
- 10. Data Models Training and Evaluation ML Topics
- 11. Challenges Getting the RIGHT data for the task And LOTs of it There is never enough data … Real World Lessons Data is crucial for successful ML projects Most boring and timeconsuming task Most underestimated task Getting the Data
- 12. Rosemary, Rosmarinus officinalis Sentiment Analysis 1245 NEGATIVE shallow , noisy and pretentious . 14575 POSITIVE one of the most splendid entertainments to emerge from the french film industry in years Iris or Flower set or example for outlier detection? 86211,B,12.18,17.84,77.79, … 862261,B,9.787,19.94,62.11, … 862485,B,11.6,12.84,74.34, … 862548,M,14.42,19.77,94.48, … 862009,B,13.45,18.3,86.6, …
- 13. Data Models Training and Evaluation ML Topics
- 14. 2012, ImageNet, G. Hinton
- 15. Data Models Training and Evaluation ML Topics
- 16. Model Complexity Training Iterationen Error Rate Training Data Test Data «Underfitting» more training needed «Overfitting» too much training model too simple model too complex
- 17. Data Models Training and Evaluation ML Topics
- 18. Supervised Learning • Learning from Examples • Right Answers are known Unsupervised Learning • Discover Structure in Data • Dimensionality Reduction Reinforcement Learning • Interaction with Dynamic Environment
- 19. Demo Time
- 20. Demo 1 Supervised Learning Pattern recognition Handwritten character recognition Convolutional neural network Demo 2 Unsupervised Learning Natural language processing (NLP) Neural word embeddings Word2vec Demos
- 21. Data Which digit is this? Collect our own data Model Deep Neural Network (LeNet-5) Deeplearning4j Deep Learning Library Open Source (Apache) Java Pattern Recognition Handwritten Digits
- 22. 1998 Gradient-based Learing for Document Recognition, Y. LeCun
- 25. Unsupervised Learning Natural Language Processing Data Google News text training dataset Texts with total of 3’000’000’000 words Lexicon: 3’000’000 words/phrases Model Word2Vec Skip-gram Mapping: Word 300-dimensional number space Many useful properties (word clustering, syntax, semantics) Deeplearning4j (Train) load and use Google News word2vec model
- 28. Recent Advances
- 29. Games Backgammon 1979, chess 1997, Jeopardy! 2011, Atari games 2014, Go 2016, Poker (Texas Hold’em) 2017 Visual CAPTCHAs 2005, face recognition 2007, traffic sign reading 2011, ImageNet 2015, lip-reading 2016 Other Age estimation from pictures 2013, personality judgement from Facebook «likes» 2014, conversational speech recognition 2016 ML performance >= Human Levels (2017) https://finnaarupnielsen.wordpress.com/2015/03/15/status-on-human-vs-machines/
- 32. 2017, Cornell + Adobe https://arxiv.org/abs/1703.07511
- 34. 2016, Erlangen, Max- Plank, Stanford http://www.graphics.stanford.edu/~niessner/papers/2016/1facetoface/thies2016face.pdf https://www.youtube.com/watch?v=ttGUiwfTYvg
- 36. Food for Thought + Next Steps
- 38. ?
- 39. Positive Outcomes Statement by Lee Sedol
- 40. Socalizing Go to talks, conferences Visit meetups (Zurich Machine Learning and Data Science, …) Increase Context Blogs, Twitter, arxiv.org, … Doing GitHub (deeplearning4j/deeplearning4j, BSI-Business-Systems-Integration-AG/anagnostes, …) Learn Python ;-) Like to learn more?
Editor's Notes
- #3: You have probably followed the story of the korean wold champion in go loosing against the alphago system built by the guys at deep mind. So, how is playing go different to playing chess from a systems perspective? This is nicely explained in the following video clip The next clip has «demis hassabis» from deepmind taking about the difference of playing go and chess from a human perspective Video 1: What is go and how does it compare to chess? Video 2: As the complexity of go is so much higher than with chess, intuition becomes even more important. Video 3: The last snippet is a high level description of the training of alpha go
- #4: But of course – as in the case of the chess playing system – there is also a hardware story behind the win of alphago --- AlphaGo was powered by TPUs in the matches against Go world champion, Lee Sedol, enabling it to "think" much faster and look farther ahead between moves. https://cloudplatform.googleblog.com/2016/05/Google-supercharges-machine-learning-tasks-with-custom-chip.html?utm_content=buffer73148&utm_medium=social&utm_source=twitter.com&utm_campaign=buffer
- #5: The Tensor Processing Units (TPU) are custom built ASIC boards that boost machine learning applications by parallelizing large amounts of low precision matrix computations. This brings around 3 generation of moore’s law making available the processing power today that we could expect to be available 6 years in the future … https://cloudplatform.googleblog.com/2016/05/Google-supercharges-machine-learning-tasks-with-custom-chip.html?utm_content=buffer73148&utm_medium=social&utm_source=twitter.com&utm_campaign=buffer
- #6: Deep reinforcement learning was used for the Alpha go A milestone article about Atari games was submitted to Nature just 2 years ago The Alpha-Go Paper https://deepmind.com/research/dqn/ http://www.nature.com/nature/journal/v518/n7540/abs/nature14236.html Deep reinforcement learning https://www.nervanasys.com/demystifying-deep-reinforcement-learning/ http://rll.berkeley.edu/deeprlcourse/ tensor flow: https://www.tensorflow.org/
- #7: https://www.nervanasys.com/demystifying-deep-reinforcement-learning/
- #8: https://www.nervanasys.com/demystifying-deep-reinforcement-learning/
- #9: https://github.com/tambetm/simple_dqn Agent implementation https://github.com/tambetm/simple_dqn/blob/master/src/agent.py DQN implementation https://github.com/tambetm/simple_dqn/blob/master/src/deepqnetwork.py https://www.nervanasys.com/openai/ https://github.com/NervanaSystems/neon
- #10: In the follwing we will Get some stories about where all this came from Talk about the major concepts Do some small demos And have a look at current work that provides a glimpse at the things to come
- #12: Most boring task and timeconsuming task of most machine learning projects (often > most of the time of a new machine learning project is used to get at the right data and enough data) Most underestimated task But, Good data sets are valuable over many decades
- #13: movie reviews: sentiment polarity http://www.cs.cornell.edu/People/pabo/movie-review-data/ Handwritten digits: mnist http://yann.lecun.com/exdb/mnist/ breast cancer: wdbc https://archive.ics.uci.edu/ml/datasets/Breast+Cancer+Wisconsin+(Diagnostic) imagenet competition http://image-net.org/challenges/LSVRC/2012/ http://image-net.org/challenges/LSVRC/2012/browse-synsets http://image-net.org/synset?wnid=n12864160 (rosemary) many more data sets http://deeplearning.net/datasets/
- #15: Massive revival with deep deep learning Classical neural networks Many layers Some tricks Rebranding from neural nets to deep learning
- #17: more training needed overtrained
- #19: http://cs229.stanford.edu/schedule.html
- #20: In the follwing we will Get some stories about where all this came from Talk about the major concepts Do some small demos And have a look at current work that provides a glimpse at the things to come
- #22: Convolutinoal nn explained http://www.wildml.com/2015/11/understanding-convolutional-neural-networks-for-nlp/?utm_content=bufferfb698&utm_medium=social&utm_source=twitter.com&utm_campaign=buffer http://yann.lecun.com/exdb/mnist/ https://deeplearning4j.org/
- #23: web online demo https://transcranial.github.io/keras-js/#/mnist-cnn
- #26: Word2vec model https://github.com/mmihaltz/word2vec-GoogleNews-vectors 3 billion words, 3 million vocabular (incl phrases such as «boston globe») http://mccormickml.com/2016/04/19/word2vec-tutorial-the-skip-gram-model/ 100 billion words from a Google News dataset Original paper (mikolov et al, 2013a) «Efficient Estimation of Word Representations in Vector Space” vector dimensionality 300 and context size 5 https://arxiv.org/pdf/1301.3781.pdf [6 billion tokens, 1 million words in voc] 2nd paper (mikolov et al, 2013b) https://arxiv.org/pdf/1310.4546.pdf []
- #30: https://finnaarupnielsen.wordpress.com/2015/03/15/status-on-human-vs-machines/ https://www.researchgate.net/publication/220271810_Computers_beat_Humans_at_Single_Character_Recognition_in_Reading_based_Human_Interaction_Proofs_HIPs
- #31: Also in 2014, researchers started to combine pattern recognition and natural language processing, both based on deep learning Stanford (2014) http://cs.stanford.edu/people/karpathy/deepimagesent/devisagen.pdf Google + stanford (2014) https://gigaom.com/2014/11/18/google-stanford-build-hybrid-neural-networks-that-can-explain-photos/
- #32: https://arxiv.org/pdf/1605.05396.pdf
- #33: https://arxiv.org/abs/1703.07511
- #34: https://research.googleblog.com/2016/09/a-neural-network-for-machine.html paper https://arxiv.org/abs/1609.08144 Newer version (multi-lingual) https://research.googleblog.com/2016/11/zero-shot-translation-with-googles.html
- #35: face2face http://www.graphics.stanford.edu/~niessner/thies2016face.html video https://www.youtube.com/watch?v=ttGUiwfTYvg paper http://www.graphics.stanford.edu/~niessner/papers/2016/1facetoface/thies2016face.pdf
- #36: Theano (python) http://deeplearning.net/software/theano/ (2011) Tensorflow (python) https://www.tensorflow.org/ (2015) Keras (Python) https://keras.io/ (2016) Pytorch (python) http://pytorch.org/ (2017) Deeplearning4j (Java) https://deeplearning4j.org/ (2014)
- #38: Financial times: https://www.ft.com/content/063c1176-d29a-11e5-969e-9d801cf5e15b Medium: https://medium.com/basic-income/deep-learning-is-going-to-teach-us-all-the-lesson-of-our-lives-jobs-are-for-machines-7c6442e37a49#.w1d0sk8dn
- #39: https://www.stlouisfed.org/on-the-economy/2016/january/jobs-involving-routine-tasks-arent-growing 1. The 90’ Computers getting used everywhere, first industrial roboters (? Need to verify!) 2. 2008 Financial crisis 3. 2017+ Future …
- #41: arXiv https://arxiv.org/list/stat.ML/current Twitter @StanfordNLP, @OpenAI @AndrewYNg, @goodfellow_ian, @karpathy, @hugo_larochelle , @ylecun , @dennybritz , @StatMLPapers