Posts Tagged ‘Artificial Intelligence’

Newer Entries »

Climatic data mining to combat global climatic issues

Tuesday, December 29th, 2009

Copenhagen summit did not only introduce the Copenhagen Accord but also a new kind of dynamics in global climate policy. The 15th United Nations Climate Change Conference (COP15) took  place at Bella Center in Copenhagen from the 7th to the 18th of December, 2009. Yes, this summit clearly brought climatic condition importance into our daily agenda, and so has data mining projects on climatic challenges have begun to rise since.

The University of Minnesota said Tuesday that it is one of the first academic partners to join the Planetary Skin Institute — a partnership between NASA and Cisco Systems Inc. that seeks to track global climate change.

The idea is to develop a global “nervous system” that will integrate land-, sea-, air- and space-based sensors. Software from University of Minnesota computer scientists will be part of the Planetary Skin prototype, set for 2010, that will track how much and where carbon is held by rain forests.

“We are excited to be an academic partner of Planetary Skin Institute,” Vipin Kumar, Researcher said. “This will allow us to greatly expedite the development and integration of advanced data-mining capabilities for the monitoring of the global ecosystem that is urgently needed in the context of climate change.”

Read more at the UM News.

Also on the other hand, A British university said Thursday it would investigate whether scientists at its prestigious Climatic Research Unit fudged data on global warming. This is a clear sign for the need of a standardized and unified approach  to solve our global climatic problems, even among scientific communities.

  • Share/Bookmark

Tags: , , ,
Posted in Data Mining, Technology, news | No Comments »

Link Compilation #5: More Video Links

Sunday, December 13th, 2009

Artificial Intelligence:

[1] Overview of AI, Agent Architectures : Video Lecture 1, Video Lecture 2

[2] Uninformed Search, Form teams : Video Lecture 3, Video Lecture 4

[3] Informed Search: Video Lecture 5

[4] Game Playing: Video Lecture 6

[5] Video Game AI, Logic: Video Lecture 8, Video Lecture 9

[6] Predicate Calculus, Inference: Video Lecture 10, Video Lecture 11

[7] Prolog Review: Video Lecture 12, Video Lecture 13

[8] Semantic Networks, Knowledge Representation: Video Lecture 14

[9] Uncertainty: Video Lecture 15, Video Lecture 16

[10] Intro Bayesian Networks, Intro Machine Learning: Video Lecture 17, Video Lecture 18

[11] Decision Trees: Video Lecture 19

[12] Neural Networks: Video Lecture 20, Video Lecture 21

[13] Genetic Algorithms, Information Retrieval: Video Lecture 22, Video Lecture 23

Natural Language Processing:

Lecture 1:

Topics:Logistics, Goals Of The Field Of NLP, Is The Problem Just Cycles?, Why NLP Is Difficult? The Hidden Structure Of Language, Why NLP Is Difficult: Newspaper Headlines, Machine Translation, Machine Translation History, Centauri/Arcturan Example

Video: Lecture 1.

Lecture 2:

Topics: Questions That Linguistics Should Answer, Machine Translation (MT), Probabilistic Language Models, Evaluation, Sparsity, Smoothing, How Much Mass To Withhold?

Video: Lecture 2

Lecture 3:

Topics: Finish Smoothing From Last Lecture, Kneser-Ney Smoothing, Practical Considerations, Machine Translation (Lecture 3), Tokenization (Or Segmentation), Statistical MT Systems, IBM Translation Models

Video: Lecture 3.

Lecture 3: http://see.stanford.edu/player/SEEslplayer.aspx?coll=63480b48-8819-4efd-8412-263f1a472f5a&co=c870ce3b-765d-4511-a070-140da17e0fdc&sl=true

Lecture 4:

Topics: Review Statistical Mt, Model 1, The Em Algorithm, Em And Hidden Structure, Em Algorithm Demonstration In Excel Spreadsheet, Assignment 1

Video: Lecture 4.

Lecture 5:

Topics: IBM Model 1-2 (Review), IBM Model 3, IBM Model 4, IBM Model 5, Mt Evaluation, Bleu Evaluation Metric, A Complete Translation System, Flaws Of Word-Based Mt, Phrased-Based Stat-Mt

Video: Lecture 5.

Lecture 6:

Topics: Continue Of Machine Translation, Syntax-Based Model, Information Extraction & Named Entity Recognition, Information Extraction, Named Entity Extraction, Precision And Recall, Naive Bayes Classifiers

Video: Lecture 6.

Lecture 7:

Topics: Continue Of Naive Bayes Classifier, Joint V.S. Conditional Models, Features, Examples, Feature-Based Classifiers, Comparison To Naïve-Bayes, Building A Maxent Model

Video: Lecture 7.

Lecture 8:

Topics: Details Of Maxent Model, Maxent Examples, Convexity, Feature Interaction, Classification, Smoothing, Inference In Systems

Video: Lecture 8.

Lecture 9:

Topics: MEMM, Hmm Pos Tagging Model, Summary Of Tagging, NER, Information Extraction And Integration, Landscape Of IE Tasks, Machine Learning Methods, Relation Extraction, Clustering

Video: Lecture 9.

Lecture 10:

Topics: Parsing, Classical NLP Parsing, Two Views Of Linguistic Structure, Attachment Ambiguities, A Simple Prediction, What Is Parsing?, Top-Down Parsing, Bottom-Up Parsing, Parsing Of PCFGs

Video: Lecture 10

Lecture 11:

Topics: Chomsky Normal Form, Cocke-Kasami-Younger (CKY) Constituency Parsing, Extended CKY Parsing, Efficient CKY Parsing, Evaluating Parsing Accuracy, How Good Are PCFGs?, Improve PCFG Parsing Via Unlexicalized Parsing, Markovization

Video: Lecture 11.

Lecture 12:

Topics: Guest Lecturer: Dan Jurafsky, Syntactic Variations Versus Semantic Roles, Some Typical Semantic Roles, Two Solutions To The Difficulty Of Defining Semantic Roles, PropBank, FrameNet, Information Extraction Versus Semantic Role Labeling, Evaluation Measures, Parsing Algorithm, Combining Identification And Classification Models, Summary

Video: Lecture 12.

Lecture 13:

Topics: Lexicalized Parsing, Parsing Via Classification Decisions: Charniak (1997), Sparseness & The Penn Treebank, Complexity Of Lexicalized PCFG Parsing, Complexity Of Lexicalized PCFG Parsing, Overview Of Collins’ Model, Choice Of Heads, The Latest Parsing Results, Parsing And Search Algorithms

Video: Lecture 13.

Lecture 14:

Topics: Parsing As Search, Agenda-Based Parsing, What Can Go Wrong?, Search In Modern Lexicalized Statistical Parsers, Dependency Parsing, Naïve Recognition/Parsing, Discriminative Parsing, Discriminative Models

Video: Lecture 14.

Lecture 15:

Topics: Why Study Computational Semantics?, Precise Semantics. An Early Example: Chat-80, Programming Language Interpreter, Logic: Some Preliminaries, Compositional Semantics, A Simple DCG Grammar With Semantics, Augmented CFG Rules, Semantic Grammars

Video:  Lecture 15.

Lecture 16:

Topics: An Introduction To Formal Computational Semantics, Database/ Knowledgebase Interfaces, Typed Lambda Calculus, Types Of Major Syntactic Categories, Adjective And PP Modification, Why Things Get More Complex, Generalized Quantifiers, Representing Proper Nouns With Quantifiers, Questions With Answers!, How Could We Learn Such Representations?

Video: Lecture 16.

Lecture 17:

Topics: Lexical Semantics, Lexical Information And NL Applications, Polysemy Vs Homonymy, WordNet, Word Sense Disambiguation, Corpora Used For WSD Work, Evaluation, Lexical Acquisition, Vector-Based Lexical Semantics, Measures Of Semantic Similarity

Video: Lecture 17.

Lecture 18:

Topics: Question Answering Systems And Textual Inference, A Brief (Academic) History, Top Performing Systems, Answer Types In State-Of-The-Art QA Systems, Semantics And Reasoning For QA, The Textual Inference Task, Why We Need Sloppy Matching, QA Beyond TREC

Video: Lecture 18.

Note: If the links either expired or exhibit errors, please visit SEE program home page and navigate from there. Thanks.

Read Link compilation #4 here.

  • Share/Bookmark

Tags: , , ,
Posted in Data Mining, Info Links, Training | 2 Comments »

Links Compilation #4: Machine Learning Videos

Saturday, December 12th, 2009

Machine Learning:

This course is provided by Stanford University SEE program. The Machine Learning course provides a broad introduction to machine learning and statistical pattern recognition. Topics include supervised learning, unsupervised learning, learning theory, reinforcement learning and adaptive control. Recent applications of machine learning, such as to robotic control, data mining, autonomous navigation, bioinformatics, speech recognition, and text and web data processing are also discussed.

Lecture 1:

Topics: The Motivation & Applications of Machine Learning, The Logistics of the Class, The Definition of Machine Learning, The Overview of Supervised Learning, The Overview of Learning Theory, The Overview of Unsupervised Learning, The Overview of Reinforcement Learning

Lecture 2:

Topics: An Application of Supervised Learning – Autonomous Deriving, ALVINN, Linear Regression, Gradient Descent, Batch Gradient Descent, Stochastic Gradient Descent (Incremental Descent), Matrix Derivative Notation for Deriving Normal Equations, Derivation of Normal Equations

Lecture 3:

Topics: The Concept of Underfitting and Overfitting, The Concept of Parametric Algorithms and Non-parametric Algorithms, Locally Weighted Regression, The Probabilistic Interpretation of Linear Regression, The motivation of Logistic Regression, Logistic Regression, Perceptron

Lecture 4:

Topics: Newton’s Method, Exponential Family, Bernoulli Example, Gaussian Example, General Linear Models (GLMs), Multinomial Example, Softmax Regression

Lecture 5:

Topics: Discriminative Algorithms, Generative Algorithms, Gaussian Discriminant Analysis (GDA), GDA and Logistic Regression, Naive Bayes, Laplace Smoothing

Lecture 6:

Topics: Multinomial Event Model, Non-linear Classifiers, Neural Network, Applications of Neural Network, Intuitions about Support Vector Machine (SVM), Notation for SVM, Functional and Geometric Margins

Lecture 7:

Topics: Optimal Margin Classifier, Lagrange Duality, Karush-Kuhn-Tucker (KKT) Conditions, SVM Dual, The Concept of Kernels

Lecture 8:

Topics: Kernels, Mercer’s Theorem, Non-linear Decision Boundaries and Soft Margin SVM, Coordinate Ascent Algorithm, The Sequential Minimization Optimization (SMO) Algorithm, Applications of SVM

Lecture 9:

Topics: Bias/variance Tradeoff, Empirical Risk Minimization (ERM), The Union Bound, Hoeffding Inequality, Uniform Convergence – The Case of Finite H, Sample Complexity Bound, Error Bound, Uniform Convergence Theorem & Corollary

Lecture 10:

Topics: Uniform Convergence – The Case of Infinite H, The Concept of ‘Shatter’ and VC Dimension, SVM Example, Model Selection, Cross Validation, Feature Selection

Lecture 11:

Topics: Bayesian Statistics and Regularization, Online Learning, Advice for Applying Machine Learning Algorithms, Debugging/fixing Learning Algorithms, Diagnostics for Bias & Variance, Optimization Algorithm Diagnostics, Diagnostic Example – Autonomous Helicopter, Error Analysis, Getting Started on a Learning Problem

Lecture 12:

Topics: The Concept of Unsupervised Learning, K-means Clustering Algorithm, K-means Algorithm, Mixtures of Gaussians and the EM Algorithm, Jensen’s Inequality, The EM Algorithm, Summary

Lecture 13:

Topics: Mixture of Gaussian, Mixture of Naive Bayes – Text clustering (EM Application), Factor Analysis, Restrictions on a Covariance Matrix, The Factor Analysis Model, EM for Factor Analysis

Lecture 14:

Topics: The Factor Analysis Model,0 EM for Factor Analysis, Principal Component Analysis (PCA), PCA as a Dimensionality Reduction Algorithm, Applications of PCA, Face Recognition by Using PCA

Lecture 15:

Topics: Latent Semantic Indexing (LSI), Singular Value Decomposition (SVD) Implementation, Independent Component Analysis (ICA), The Application of ICA, Cumulative Distribution Function (CDF), ICA Algorithm, The Applications of ICA

Lecture 16:

Topics: Applications of Reinforcement Learning, Markov Decision Process (MDP), Defining Value & Policy Functions, Value Function, Optimal Value Function, Value Iteration, Policy Iteration

Lecture 17:

Topics: Generalization to Continuous States, Discretization & Curse of Dimensionality, Models/Simulators, Fitted Value Iteration, Finding Optimal Policy

Lecture 18:

Topics: State-action Rewards, Finite Horizon MDPs, The Concept of Dynamical Systems, Examples of Dynamical Models, Linear Quadratic Regulation (LQR), Linearizing a Non-Linear Model, Computing Rewards, Riccati Equation

Lecture 19:

Topics: Advice for Applying Machine Learning, Debugging Reinforcement Learning (RL) Algorithm, Linear Quadratic Regularization (LQR), Differential Dynamic Programming (DDP), Kalman Filter & Linear Quadratic Gaussian (LQG), Predict/update Steps of Kalman Filter, Linear Quadratic Gaussian (LQG)

Lecture 20:

Topics: Partially Observable MDPs (POMDPs), Policy Search, Reinforce Algorithm, Pegasus Algorithm, Pegasus Policy Search, Applications of Reinforcement Learning

Reader Older Link Compilation #3 here.

  • Share/Bookmark

Tags: , , , ,
Posted in Data Mining, Info Links, Machine Learning, Research, Training | 1 Comment »

E-Memory Revolution. Are we prepared?

Monday, September 21st, 2009

Creating E-Memories, remembering everything and retrieving them in a click is no more a part of a science friction movie. It’s possible in today’s world. The advent of this technology lead to the necessity to store and retrieve. Not only would our daily activities get recorded, but also certain personal information such as reading habits, health, location, and computer usage be recorded with little or no effort. To phrase it in simple terms, future library will have digital library section that will have digital details about the person, in terms of whatever he has read, watched and heard so far. There are three pioneer technologies involved in this particular “Total Recall” revolution. The first is recording technology, which utilizes digital cameras and cell phones that include cameras. Their functionalities include location-tracking, environment-sensing (for example, temperature and humidity), and biometric sensing (via on-body devices today and someday via in-body devices). Furthermore, the traces of one’s digital transactions can tell a detailed story—whose screenplay does not include itemized credit card bills but phone call logs, email in-boxes, web browsing histories, movie rentals alone, but much more. The digital records that we gather is much more than we actually imagine. The second technology stream is the rapid increase in capacity and corresponding decrease in price for digital storage. Imagine an archive of everything you ever read or wrote—books, articles, web pages, emails, letters, and so on—audio and video included. You could now possibly imagine the amount of space it would take.  As years roll by, this data can fit a cell phone or be consumed by a cell phone. A few years after that, rolling video nonstop throughout life will be possible. The third stream is powerful software to take advantage of a lifetime of e-memories. What Google does, that is, searching for words in web pages, is the crux of this software.  It will be possible to find things by cross-correlation, such as the document you read while in Phoenix, the picture taken by a relative, or the email you sent on that particularly cold day. And data-mining software will crunch through your life-log, finding patterns, trends, and connections. This third stream is where the most dramatic developments are occurring now. Having a foresight, this ideology may be useful in various possible ways. Tracking down to the heredity disease that originated from your great-grandfather is one such benefit. Thinking in a broader sense, not every one has the possibility of recording video and audio over internet. So what do we actually do with the records?! The point blank answer that is possible is, “Prediction of the behavior”. When I say, prediction, it is as intelligent as predicting the response in a dialogue.

Researchers at Carnegie Mellon University, Pittsburgh, have a program that lets you ask questions of Albert Einstein. You can chat with a virtual George Bush or Bart Simpson at MyCyberTwin.com.

deploy-website

The advent of e-books has almost become inevitable for every student. So why do I emphasize about the e-book in the e-memory article. The reason is that, it opens the next level of e-memory, which has logs about lifelong learning details.  It can enable electronic highlighting and note-taking. It will record lab experiments and conversations with the teacher. Increasing use of electronic book readers and the ubiquity of notebook computers will propel the ascendency of e-texts, which will open the door to a new world of learning with Total Recall.

For example, in a family, mother may browse over certain books and articles relating to her interest and son may roll over certain sports columns in websites. Likewise the tastes and opinions differ in a family. So, a portal with a record of their e-memory activities will enable the portal to harness their likes and dislikes and display the products accordingly. This opens up a new horizon for marketing benefits.

Speaking of digital achieves, data loss, data decay, and data entanglement are major points of concern. Backup and replication techniques now have made data loss a thing of the past. Data decay and data entanglement are more problematic, but with proper technology in place these can be avoided.

Data decay occurs when the format you stored something in becomes obsolete and unreadable. Putting it in a different sense, it is simple as storing a print version of an old spreadsheet file. Though we cannot perform any further calculations as in case of a spreadsheet file, it is safer to have a backup of some sort. The benefits of e-memories will extend across the life of the individual and throughout society. These benefits, along with the technological trends that make e-memories affordable and convenient, make the Total Recall era inevitable. Your life, in so far as it is information, is about to become totally accessible to you.

-Vidhya, Student Intern

  • Share/Bookmark

Tags: ,
Posted in Research, Technology, Tools | No Comments »

Newer Entries »