Summary: Four proposed metrics:
[1] average relative reduction in training time (sample size, number of training experiences)
[2] jumpstart (initial advantage of transfer algorithm)
[3] handicap (how long it takes the no-transfer algorithm to overcome the jumpstart)
[4] asymptotic advantage (how much better the transfer learning algorithm does in the...
Although machine learning is becoming commonly used in today's software, there has been little research into how end users might interact with machine learning systems, beyond communicating simple "right/wrong" judgments. If the users themselves could somehow work hand-in-hand with machine learning systems, the accuracy of learning systems could be improved...
The field of machine learning has made major strides over the last 20 years. This document summarizes the major problem formulations that the discipline has studied, then reviews three tasks in cognitive networking and briefly discusses how aspects of those tasks fit these formulations. After this, it discusses challenges for...
This paper studies the problem of learning diagnostic policies from training examples. A diagnostic policy is a complete description of the decision-making actions of a diagnostician (i.e., tests followed by a diagnostic decision) for all possible combinations of test results. An optimal diagnostic policy is one that minimizes the expected...
The standard model of supervised learning assumes that training and test data are drawn from the same underlying distribution. This paper explores an application in which a second, auxiliary, source of data is available drawn from a different distribution. This auxiliary data is more plentiful, but of significantly lower quality,...
This paper addresses cost-sensitive classification in the setting where there are costs for measuring each attribute as well as costs for misclassification errors. We show how to formulate this as a Markov Decision Process in which the transition model is learned from the training data. Specifically we assume a set...
A common heuristic for solving Partially Observable Markov Decision Problems POMDPs is to first solve the underlying Markov Decision Process MDP and then construct a POMDP policy by performing a fixed depth lookahead search in the POMDP and evaluating the leaf nodes using the MDP value function. A problem with...
This paper introduces the even-odd POMDP an approximation to POMDPs Partially Observable Markov Decision Problems in which the world is assumed to be fully observable every other time step. This approximation works well for problems with a delayed need to observe. The even-odd POMDP can be converted into an equivalent...
What is the relationship between learning and reasoning? Much recent work in machine learning has been criticized for focusing on learning and ignoring reasoning. This paper attempts to describe the various ways in which machine learning research has (and has not) incorporated reasoning. The paper argues that there are important...
This paper presents qualitative results from interviews with knowledge workers about their recovery strategies after interruptions. Special focus is given to when these strategies fail due to the nature of the interruption and existing computer support. Potential solutions offered by participants to overcome some of these problems are presented. These...
Many machine learning applications require
classifiers that minimize an asymmetric cost
function rather than the misclassification
rate, and several recent papers have addressed
this problem. However, these papers
have either applied no statistical testing
or have applied statistical methods that are
not appropriate for the cost-sensitive setting.
Without good statistical...
This paper introduces the even-odd POMDP, an approximation to POMDPs in which the world is assumed to be fully observable every other time step. The even-odd POMDP can be converted into an equivalent MDP, the
2MDP, whose value function, V*[subscript 2MDP], can be combined online with a 2-step lookahead search...