Presentation is loading. Please wait.

Presentation is loading. Please wait.

Machine learning techniques and Applications

Similar presentations


Presentation on theme: "Machine learning techniques and Applications"— Presentation transcript:

1 Machine learning techniques and Applications
Part I Mahdi Eftekhari Associate Professor in Department of Computer Engineering, Shahid Bahonar University of Kerman, Kerman, Iran Last Update: Dec 2016

2 Contents Introduction Taxonomy of Learning Methods
Supervised learning Applications Unsupervised Learning Applications Reinforcement Learning Applications Large Scale machine learning Future trends Machine learning techniques and Applications M.Eftekhari

3 Introduction Introduction Taxonomy of Learning Methods Supervised learning Applications Unsupervised Learning Applications Reinforcement Learning Applications Large Scale machine learning Future trends Machine learning is programming computers to optimize a performance criterion for the agent using example data or past experience. Agent Examples: Humanoid robots, computer programs (chatbots, softbots, minerbots etc….) Learning is used for making the agent autonomous Data for learning past Experience Explicit data  training and learning models based on data Implicit data  The learning agent must try in environment for gathering the data and experience Machine learning techniques and Applications M.Eftekhari 3

4 Taxonomy of Learning Methods
Introduction Taxonomy of Learning Methods Supervised learning Applications Unsupervised Learning Applications Reinforcement Learning Applications Large Scale machine learning Future trends Supervised Learning (Explicit data) Classification Regression/Prediction Unsupervised Learning (Explicit data) Reinforcement Learning(implicit data) Many disciplines Like: Statistic, Computer Science, Mathematics, Robotics, Psychology, Cognitive science. Machine learning techniques and Applications M.Eftekhari 4

5 Supervised learning (classification)
Introduction Taxonomy of Learning Methods Supervised learning Applications Unsupervised Learning Applications Reinforcement Learning Applications Large Scale machine learning Future trends Example: Credit scoring Differentiating between low-risk and high-risk customers from their income and savings Discriminant model: IF income > θ1 AND savings > θ2 THEN low-risk ELSE high-risk Machine learning techniques and Applications M.Eftekhari 5

6 Supervised learning (classification Applications)
Introduction Taxonomy of Learning Methods Supervised learning Applications Unsupervised Learning Applications Reinforcement Learning Applications Large Scale machine learning Future trends Face recognition: Pose, lighting, make-up, hair style Character recognition: Different handwriting styles. Speech recognition: Sensor fusion: Combine multiple modalities; eg, visual (lip image) and acoustic for speech Medical diagnosis: From symptoms to illnesses Web Advertising: Predict if a user clicks on an advertise link on the Internet. SPAM detection and Filtering. Machine learning techniques and Applications M.Eftekhari 6

7 Supervised learning (classification Applications)
Introduction Taxonomy of Learning Methods Supervised learning Applications Unsupervised Learning Applications Reinforcement Learning Applications Large Scale machine learning Future trends Example: decision trees that create rules Prediction of future cases: Use the rule to predict the output for future inputs Knowledge extraction: The rule is easy to understand Outlier detection: Exceptions that are not covered by the rule, e.g., fraud Develop a prevention mechanism against online banking cyber-crime using machine learning (ML). Other classifiers: Bayesian classifiers, Artificial Neural Networks, Fuzzy Rule based classifiers, k-Nearest Neighbor approaches, ….etc. Machine learning techniques and Applications 7 M.Eftekhari 7

8 Supervised learning (Regression)
Introduction Taxonomy of Learning Methods Supervised learning Applications Unsupervised Learning Applications Reinforcement Learning Applications Large Scale machine learning Future trends Example: Price of a used car x : car attributes y : price y = g (x ; θ ) g ( ) model, θ parameters Machine learning techniques and Applications M.Eftekhari 8

9 Supervised learning (Regression)
Introduction Taxonomy of Learning Methods Supervised learning Applications Unsupervised Learning Applications Reinforcement Learning Applications Large Scale machine learning Future trends Estimating the quality of VOIP (Voice Over IP) Navigating a car: Angle of the steering wheel Kinematics of a robot arm Estimating the performance of CPU. Neural networks, regression trees, least square estimators, Support vector regression, genetic programming, Fuzzy Inference Systems, …etc. α1 α2 (x,y) α1= g1(x,y) α2= g2(x,y) Machine learning techniques and Applications M.Eftekhari 8

10 Unsupervised learning
Introduction Taxonomy of Learning Methods Supervised learning Applications Unsupervised Learning Applications Reinforcement Learning Applications Large Scale machine learning Future trends Clustering: Grouping similar instances Other applications: text Summarization, Association Analysis Example applications Customer segmentation in Customer relationship management (CRM). Image segmentation. Separating different sources of voice in a mixed recording voice (ICA independent component analysis). Recommender systems: Recommending similar people in face book, similar books in amazon, similar videos ,…. Machine learning techniques and Applications M.Eftekhari 8

11 Reinforcement Learning
Introduction Taxonomy of Learning Methods Supervised learning Applications Unsupervised Learning Applications Reinforcement Learning Applications Large Scale machine learning Future trends Topics: Policies: what actions should an agent take in a particular situation Utility estimation: how good is a state (used by policy) No supervised output but delayed reward Credit assignment problem (what was responsible for the outcome) Machine learning techniques and Applications M.Eftekhari 8

12 Reinforcement Learning
Introduction Taxonomy of Learning Methods Supervised learning Applications Unsupervised Learning Applications Reinforcement Learning Applications Large Scale machine learning Future trends Applications: Game playing Robot in a maze Multiple agents, partial observability, ... Learning Classifier Systems (online Intrusion Detection and Protection Systems (IDS,IPS)) Web page ranking (online ranking via relevance feedback) Learning hard maneuvers. Learning Chatbots. Machine learning techniques and Applications M.Eftekhari 8

13 Large Scale machine learning
Introduction Taxonomy of Learning Methods Supervised learning Applications Unsupervised Learning Applications Reinforcement Learning Applications Large Scale machine learning Future trends In particular, existing algorithms with linear complexity are computationally infeasible in the context of millions or even billions of data points. Massive data: Networks’ logs, Fly information as massive streaming data, bioinformatics, Social networks. Big data and large scale data. Adapting traditional machine learning approaches to deal with time and space challenges. Deep learning Machine learning techniques and Applications M.Eftekhari 8

14 Large Scale machine learning (tools)
Introduction Taxonomy of Learning Methods Supervised learning Applications Unsupervised Learning Applications Reinforcement Learning Applications Large Scale machine learning Future trends Map-Reduce for Machine Learning on Multicore A MapReduce job usually splits the input data-set into independent chunks which are processed by the map tasks in a completely parallel manner. The framework sorts the outputs of the maps, which are then input to the reduce tasks. Tensor Flow : . It maps the nodes of a dataflow graph across many machines in a cluster, and within a machine across multiple computational devices, including multicore CPUs, general purpose GPUs, and custom-designed ASICs known as Tensor Processing Units (TPUs). Machine learning techniques and Applications M.Eftekhari 8

15 Large Scale machine learning (tools)
Introduction Taxonomy of Learning Methods Supervised learning Applications Unsupervised Learning Applications Reinforcement Learning Applications Large Scale machine learning Future trends  Developing a software framework called DistBelief that can utilize computing clusters with thousands of machines to train large models. J. Dean, G.S. Corrado, R. Monga, K. Chen, M. Devin, Q.V. Le, M.Z. Mao, M.A. Ranzato, A. Senior, P. Tucker, K. Yang, A. Y. Ng in NIPS 2012. VFML (Very Fast Machine Learning) toolkit for mining high-speed data streams and very large data sets Machine learning techniques and Applications M.Eftekhari 8

16 Large Scale machine learning
Introduction Taxonomy of Learning Methods Supervised learning Applications Unsupervised Learning Applications Reinforcement Learning Applications Large Scale machine learning Future trends P. Domingos and G. Hulten, Mining High-Speed Data Streams. Proceedings of the Sixth International Conference on Knowledge Discovery and Data Mining (pp ), Boston, MA: ACM Press. P. Domingos and G. Hulten, A General Method for Scaling Up Machine Learning Algorithms and its Application to Clustering. Proceedings of the Eighteenth International Conference on Machine Learning (pp ), Williamstown, MA: Morgan Kaufmann. P. Domingos and G. Hulten, Learning from Infinite Data in Finite Time. Advances in Neural Information Processing Systems 14 (pp ), Cambridge, MA: MIT Press. Machine learning techniques and Applications M.Eftekhari 8

17 Future trends Introduction Taxonomy of Learning Methods
Supervised learning Applications Unsupervised Learning Applications Reinforcement Learning Applications Large Scale machine learning Future trends Machine learning techniques and Applications M.Eftekhari 8

18 Future trends Introduction Taxonomy of Learning Methods Supervised learning Applications Unsupervised Learning Applications Reinforcement Learning Applications Large Scale machine learning Future trends Google Photos, which uses a neural network-based image recognition system to automatically classify photos by their contents. Facebook’s system for tagging people in photos Machine learning techniques and Applications M.Eftekhari 8

19 Future trends Introduction Taxonomy of Learning Methods Supervised learning Applications Unsupervised Learning Applications Reinforcement Learning Applications Large Scale machine learning Future trends Artificial Narrow Intelligence (ANI):  Artificial Narrow Intelligence is AI that specializes in one area (we are now). Artificial General Intelligence (AGI): Sometimes referred to as Strong AI, or Human-Level AI—a machine that can perform any intellectual task that a human being can. Creating AGI is a much harder task than creating ANI Artificial Superintelligence (ASI): superintelligence as “an intellect that is much smarter than the best human brains in practically every field, including scientific creativity, general wisdom and social skills.” Machine learning techniques and Applications M.Eftekhari 8

20 Future trends Introduction Taxonomy of Learning Methods
Supervised learning Applications Unsupervised Learning Applications Reinforcement Learning Applications Large Scale machine learning Future trends Machine learning techniques and Applications M.Eftekhari 8

21 Pushing machine learning onto mobile devices
Future trends Introduction Taxonomy of Learning Methods Supervised learning Applications Unsupervised Learning Applications Reinforcement Learning Applications Large Scale machine learning Future trends Pushing machine learning onto mobile devices Research efforts at MIT and IBM suggest that we will soon see more chips on the market that excel at running neural networks at high speed, in small spaces and at low power. Performing sophisticated feats that take advantage of neural networks, such as computer vision and speech recognition Machine learning techniques and Applications M.Eftekhari 8

22 Future trends Introduction Taxonomy of Learning Methods Supervised learning Applications Unsupervised Learning Applications Reinforcement Learning Applications Large Scale machine learning Future trends Small Size Hardware modules dedicated to machine learning NVIDIA software toolkits for accelerated computing, deep learning, computer vision, and graphics. Machine learning techniques and Applications M.Eftekhari 8

23 Future trends Introduction Taxonomy of Learning Methods Supervised learning Applications Unsupervised Learning Applications Reinforcement Learning Applications Large Scale machine learning Future trends praised the creation of an academic institute dedicated to researching the future of intelligence as “crucial to the future of our civilization and our species”. Machine learning techniques and Applications M.Eftekhari 8


Download ppt "Machine learning techniques and Applications"

Similar presentations


Ads by Google