Sunday, August 26, 2007

Lazy Learning

In artificial intelligence, lazy learning is a learning method in which generalization beyond the training data is delayed until a query is made to the system, as opposed to in eager learning, where the system tries to generalize the training data before receiving queries.


Hausdorff distance

what is Hausdorff distance ?
An introduction

Named after Felix Hausdorff (1868-1942), Hausdorff distance is the « maximum distance of a set to the nearest point in the other set »

Sunday, August 19, 2007

A talk given by Prof. Mitch Marcus on last Friday

> タイトル:Unsupervised induction of morphological structure
> 概要:
> We will discuss the problem of unsupervised morphological and part of
> speech (POS) acquisition in realistic settings. From studies of tagged
> corpora, we show that there is a sparse data problem in morphology,
> which raises the question of how rare forms may be learned. We then show
> that it is often the case that the base form of a word is present among
> the different inflections of a lexeme, which suggests that rare forms
> can be learned by association with a base form. We introduce new
> representations for morphological structure which express the
> morphophonological transduction behavior of these base forms, and
> present
> an algorithm to acquire these structures automatically from an unlabeled
> corpus. We apply the algorithm to a range of Indo-European languages
> including Slovene, English, and Spanish.

1. met the same group of people (well, I mean young researchers basically) again.
2. I asked two questions on how to deal with sparse data. Based on what I understood:
a) To prune the space by analyzing features of the data.
b) To add back ground knowledge.
3. Jin said he is the 牛魔王 in their field! Orz...

Friday, August 17, 2007



\centering \subfigure[subCaption_1 ]{\includegraphics[width=200pt]{1.eps}}
\label{fig:selFilter} \subfigure[subCaption_2]{\includegraphics[width=200pt]{2.eps}}
\caption{Query} \label{Fig:CellDropRates}

Tuesday, August 7, 2007

Maxium Likelihood Estimate (MLE)

1. from wikipedia:

Maximum likelihood estimation (MLE) is a popular statistical method used to make inferences about parameters of the underlying probability distribution from a given data set. That is to say, you have a sample of data

X_{1}, \dots, X_{n} \!

and some kind of model for data, and you want to estimate parameters of the distribution.

2. from
Maximum likelihood estimation begins with the mathematical expression known as a likelihood function of the sample data. Loosely speaking, the likelihood of a set of data is the probability of obtaining that particular set of data given the chosen probability model. This expression contains the unknown parameters. Those values of the parameter that maximize the sample likelihood are known as the maximum likelihood estimates.

the dis/advantages are discussed, as well as the software.

3. about smoothed maximum likelihood estimates.
One purpose of the smoothed estimates is too account for sparseness in counts for distributions with a lot of history by backing off to less sparse estimates.
(McDonald, R. (2005). Extracting Relations from Unstructured Text, Department of Computer and Information Science, University of Pennsylvania.)