2015年09月_zypandora

原创 Lecture5-3Effective number of hypotheses

Dichotomies={hypothesis h:→{x,o}}\mathcal{H} = \{\textrm{hypothesis } h: \mathcal{X} \rightarrow \{x,o\}\}Call: h(x1,x2,...,xN)=(h(x1),h(x2),...,h(xN))∈{x,o}Nh(\mathbf{x_1, x_2, ... , x_N}) = (h(\m

2015-09-29 22:49:41 498

原创 Lecture5-2Effective number of lines

Where did MM come from?Where did Union Bound fail? Union Bound Over-Estimating!! Can we group hypotheses by kind?How many lines are there?For only one point x1x_1, how many different lines are t

2015-09-29 22:16:08 277

原创 Lecture5-1Recap and Preview

Flow Chart Recap2 Central QuestionsLearning is split into 2 questions - Can we make sure that Ein≈EoutE_{in} \approx E_{out}? - Can we make EinE_{in} small enough?Trade off on MMSmall MM :) few ch

2015-09-27 21:47:06 202

原创 Lecture4-4Connection to Real Learning

Multiple hhReal Learning like PLA, what about when getting all green(right)?Bad Sample:EinE_{in} and EoutE_{out} far away – can get worse results. Eout=12E_{out} = \frac{1}{2}, but getting all heads(

2015-09-22 21:57:48 203

原创 Lecture4-3Connection to Learning

Bin Model vs. Learningunknown orange prob μ\mu ↔\leftrightarrow fixed hypothesis h(x)=?h(\mathbf{x}) =? target f(x)f(\mathbf{x})marbel ∈\in bin ↔\leftrightarrow x∈\mathbf{x} \in \mathcal{X}orange ↔

2015-09-20 22:32:00 155

原创 Lecture4-2Probability to rescure

Inferring something unknownDifficult to infer unknown target ff outside \mathcal{D}, can we infer something unknown in other scenarios?How to infer the orange probability?SamplingBin: ASSUME orange p

2015-09-20 13:26:02 198 1

原创 Lecture4-1Learning is impossible?

Puzzle problemNo free lunchWe all have to be aware that ff is unknown, and any possible ff can happen according to the previous examples outside \mathcal{D}(no free lunch). So we have to make some ass

2015-09-20 12:28:07 251

原创 Lecture3-3Learning with different input space

Concrete featureRaw featureAbstract featureNeed to extract real concrete features from abstract features Mini SummarySummary

2015-09-20 12:10:21 213

原创 Lecture3-3Learning with different protocols

Batch LearningOnline LearningActive LearningMini Summary

2015-09-20 11:54:35 256

原创 Lecture3-1Learning with different data label

Supervised LearningUnsupervised Learning Semi-supervised LearningReinforcement LearningA very different but natural way of learningTeach your dog: SIT DOWNcannot easily show the dog that yn=y_n = sit

2015-09-20 11:36:59 197

原创 Lecture3-1Learning with different output space

Binary ClassificationMulticlass Classification: Coin Recognition ProblemRegression: Patient Recovery Prediction ProblemStructured Learning: TaggingMini Summary

2015-09-20 10:45:08 349

原创 Lecture2-4Non-Separable Data

In reality, we don’t know if \mathcal{D} is linear separable, and we don’t know what T exactly is, because we cannot calculate ρ\rho, which depends on wf\mathbf{w_f}.Flow Chart 2: With Noise What abou

2015-09-19 23:21:41 232

原创 Lecture2-3Guarantee of PLA

Linear SeparableIf PLA halts(no mistakes),(necessary condition) \mathcal{D} allows some w\mathbf{w} to make no mistakeCall \mathcal{D} linear separable Linear Separable ⟺\mathcal{D} \Longleftrigh

2015-09-19 22:11:03 318

原创 Lecture2-2Perceptron Learning Algorithm

Select gg from \mathcal{H}=\mathcal{H}=all possible perceptrons, g=?g = ?want g≈fg \approx falmost necessary: g≈fg \approx f on \mathcal{D}, ideally g(xn)=f(xn)=yng(x_n)=f(x_n)=y_nDifficult: \m

2015-09-19 18:30:07 317

原创 Lecture2-1Perceptron Hypothesis Set

Credit Approval RevisitedA simple hypothesis set: PerceptronFor x=(x1,x2,...,xd)\mathbf{x} = (x_1, x_2, ..., x_d), ‘features of customer’, compute a weight ‘score’ and - approve credit if ∑di=1wixi>th

2015-09-15 22:37:17 260

原创 U1-Introduction

Anecdotal EvidenceAssumption that may be real but is only based on limited number of samples, which is not representative. Population and samplesKeysPopulationSampleDesignScopeExploratory data an

2015-09-14 21:36:46 132

原创 Week1-11Memory

Basics

2015-09-13 12:16:20 255

原创 week1-10Theory of algorithms

Type of analysesBest case: lower boundWorst case: upper boundAverage case: expected cost for random inputActual data might not match input model?Need to understand the input to effectively process

2015-09-13 12:04:16 184

原创 Week1-9Order-of-Growth classifications

Common Order-of-Growth classificationsSmall set of functions 1,logN,N,NlogN,N2,N3,abN1, logN, N, NlogN, N^2, N^3, ab^Nlinearithmic and linear scale with the input size!!Binary SearchImplementationpu

2015-09-13 11:31:50 437

原创 Week1-8Mathematical Model

Mathematical model for running timeCost of basic operations And there are many many other ops that can be calculated.Example:1-SUM2-SUM Maybe we should just count the ones that are most expensive!!!

2015-09-13 00:03:39 253

原创 Lecture 1-5ML and Other Fields

ML and Data MiningMachine Learning Use data to compute hypothesis gg that approximates target ffText Mning Use (huge) data to find property that is interestingif ‘interesting property’ same as ‘hy

2015-09-12 23:47:57 259

原创 Lecture1-4Components of ML

Components of ML: Metaphor using credit approvalThe bank has the applicant information, and there is unknown pattern to be learned, which is used to solve the final question: whether approving credit

2015-09-12 23:30:41 229

原创 week1-7Observations

Example: 3-SUMGiven N distinct integers, how many triples sum to exactly zero?Brute Force AlgorithmMeasuring the running timeWe can calculate the real running time for various input sizes and make a pl

2015-09-12 17:32:40 236 1

原创 Week1-6Analysis of Algorithm Introduction

Running Time

2015-09-12 17:07:55 281

原创 Week1-5Union-Find Applications

ApplicationsPercolationlikelihood of percolationDepends on the site vacancy probability p.Percolation phase transition No analytic solutions, only computer simulation answers!!Solution: Monte Carlo

2015-09-12 16:16:36 192

原创 Week1-4Qucik-Union Improvments

Improvement 1: weightingweighted quick unionmodify quick-union to avoid tall treeskeep track of size of each tree(number of each objects)balance by linking root of smaller tree to the root of larger

2015-09-11 23:22:38 244

原创 Lecture1-3Applications of ML

Applications of daily needsFooddata: twitter data(words + location)skill: tell food poisoning likeliness of restaurant properlyClothing-data: sales figures + client surveys -skill: give good fashio

2015-09-11 23:13:10 239

原创 Lecture1-2What is Machine Learning?

From Learning to Machine Learning

2015-09-11 22:44:26 510

原创 Week1-3Quick Union

Data StructureInteger array id[] of size NRoot of i isid[id[id[…id[i]…]]](keep going until it doesn’t change)OperationsFind( p, q ): check if p and q have the same rootUnion( p, q ): to merge comp

2015-09-09 22:32:15 246

原创 Lecture1-1Course Introduction

Course Design

2015-09-09 22:09:19 173

原创 Week1-2Quick Find

Quick Find(eager approach)Data Structure-Integer Array id[] of size N -Interpretation: p and q are connected iff they have the same idExample Findconnected( p, q )Check if p and q have the same idUni

2015-09-05 17:05:09 207

原创 Week1-1Dynamic Connectivity

Given a set of N objects, main operations are:-Union Command: Connect 2 objects: union( a, b ) -Find/Connected query: is there a path connecting the 2 objects? connect ( a, b )Answer the question if t

2015-09-05 16:12:49 234

原创 Week0-1Course Overview

Focus-Algorithm -Data Structure The course is separated into 2 parts. Why study algorithm?-To solve problems that could not be otherwise addressed. -For intellectual stimulation. -To become a prof

2015-09-04 23:29:55 236

zypandora的博客