Model Selection and Multi-Model Inference

Model Selection and Multi-Model Inference

Whilst reading the String Distance Metrics paper that Jose posted, I had lots of thoughts going through head about some current work I'm doing. In my mind, part of all this relates to modelling data sets and things of that nature. I'm HORRIBLE at that... for now. The link above is not a paper, but a link to a book on Amazon. Essentially it goes through and talks about methods used to take inputs of a "good" scientific question and your data set and come up with a list of models that could be used to help you realize what you've got. It goes over Akaike's Information Criterion and things of that nature. I am definitely NOT a stats or modelling guy, so I figured this book would be of use to anyone who is starting to deal with large data sets and need to learn some things as to how to approach handling all that data.

Language Trees and Zipping

This paper presents a general approach to comparing two sets of information:
http://arxiv.org/abs/cond-mat/0108530.
They focus on strings of text, but the technique is quite powerful and general.

x86 Binary Reoptimization

http://www.crhc.uiuc.edu/IMPACT/ftp/report/impact-98-05.binary.pdf

"An Overview of the IMPACT x86 Binary Reoptimization Framework" by Matthew C. Merten and Michael S. Thiems. This, I think, is from 1998, so a bit old. Just found it interesting -- are there any tools like this for ELF format?

A Comparison of String Distance Metrics for Name-Matching Tasks

I've been looking at a lot of soft string matching metrics lately, this paper and the SecondString library have been interesting. A related project is MinorThird, "a collection of Java classes for storing text, annotating text, and learning to extract entities and categorize text." And finally, see Maxent, from a different group, described as a "Java package for training and using maximum entropy models." Too bad it's in Java ...

For more info on their methods and library, see their paper:

Using an open-source, Java toolkit of name-matching methods, we experimentally compare string distance metrics on the task of matching entity names. We investigate a number of different metrics proposed by different communities, including edit-distance metrics, fast heuristic string comparators , token-based distance metrics, and hybrid methods. Overall, the best-performing method is a hybrid scheme combining a TFIDF weighting scheme, which is widely used in information retrieval, with the Jaro-Winkler string-distance scheme, which was developed in the probabilistic record linkage community.

Source: A Comparison of String Distance Metrics for Name-Matching Tasks William W. Cohen, Pradeep Ravikumar, and Stephen E. Fienberg.

Permissive Action Links

http://www1.cs.columbia.edu/~smb/nsam-160/pal.html

A PAL -- a "Permissive Action Link" -- is the box that is supposed to prevent unauthorized use of a nuclear weapon. "Unauthorized" covers a wide range of sin, from terrorists who have stolen bombs to insane American military officers to our allies who may have some of their own uses for bombs that are covered by joint use agreements. It's supposed to be impossible to "hot-wire" a nuclear weapon. Is it?

Plan9 -- Reference documentation

High level plan9 docs.

http://whatexit.org/tal/mywritings/plan9_vs_telnet.pdf.

Newsham, North, and many others (Mike C. from xprime@#!#$) have been playing with p9 for awhile, so figured I'd put up links.

Win2k Loader "internals" paper

Been doing some research in this area so I figured I'd shoot it across. I haven't read it yet, so could be crap (sorry). But essentially discusses some of the DLL loading mechanisms.

What goes on inside Windows 2000: Solving the mysteries of the loader -- Russ Osterlund.

The Human Immune System and Network Intrusion Detection

This paper is pretty basic, but I figured I'd post just to throw some thoughts out there. A few years ago I had been really into reading medical papers on the human immune system... stepping back into that realm.

immune1.pdf

Not much..

Data Tastes Better Seasoned: Introducing the ASH Family of Hashing Algorithms

I'm no cryptographer, but this looks interesting.

Over the recent months it has become clear that the current generation of cryptographic hashing algorithms are insufficient to meet future needs. The ASH family of algorithms provides modifications to the existing SHA-2 family. These modifications are designed with two main goals: 1) Providing increased collision resistance. 2) Increasing mitigation of security risks post-collision. The unique public/private sections and salt/pepper design elements provide increased flexibility for a broad range of applications. The ASH family is a new generation of cryptographic hashing algorithms.

Source: Data Tastes Better Seasoned: Introducing the ASH Family of Hashing Algorithms, D.J. Capelis (via arXiv).

Silver Bullet

http://users.adelphia.net/~lilavois/Cosas/Reliability.htm.
Silver bullet to fix software reliability and scalability.

NCryptFS: A Secure and Convenient Cryptographic File System

Since I was reading the GBDE paper and I believe silitek was playing with NCryptFS, I figured I should read the paper. I've note yet read, so ... don't blame me! :-) J/k

Link to paper

Data Communications: The first 2500 years.

G. Holzmann
Data
Communications: The first 2500 years

Yes, its all old-hat.

GBDE - GEOM Based Disk Encryption

FreeBSD based, but was told by others that there was interest in porting to plan9. Since I had browsed this paper awhile ago but never read ... and there was porting interest, figured I'd re-read. Here's the link:


GBDE - Poul-Henning Kamp