Hypothesizing an algorithm from one example: the role of specificity.

Statistical machine learning usually achieves high-accuracy models by employing tens of thousands of examples. By contrast, both children and adult humans typically learn new concepts from either one or a small number of instances. The high data efficiency of human learning is not easily explained in terms of standard formal frameworks for machine learning, including Gold's learning-in-the-limit framework and Valiant's probably approximately correct (PAC) model. This paper explores ways in which this apparent disparity between human and machine learning can be reconciled by considering algorithms involving a preference for specificity combined with program minimality. It is shown how this can be efficiently enacted using hierarchical search based on identification of certificates and push-down automata to support hypothesizing compactly expressed maximal efficiency algorithms. Early results of a new system called DeepLog indicate that such approaches can support efficient top-down construction of relatively complex logic programs from a single example. This article is part of a discussion meeting issue 'Cognitive artificial intelligence'.

Outcome of a controlled trial of the effectiveness of intensive case management for chronic public inebriates.

OBJECTIVE: The objective of this study was to test whether an intensive case management intervention would be effective with a group of homeless chronic public inebriate clients. The primary goals of the case management were to improve the financial and residential stability of the clients and to reduce their use of alcohol. METHOD: Subjects (N = 298, 81% male) were interviewed at baseline, randomly assigned to treatment and control conditions and given follow-up interviews at 6-month intervals for 2 years. Case management services were provided for the duration of the project. Follow-up rates for the first three interviews averaged 82%. RESULTS: Repeated measures MANCOVAs showed significant group differences favoring the case-managed group in all three areas targeted by the intervention: total income from public sources, nights spent in "own place" out of the previous 60 nights and days drinking out of the previous 30 days. The results held whether the three variables were analyzed jointly or separately and for alternative measures of drinking and homelessness. Although statistically significant, the group differences are generally not large. CONCLUSIONS: The results indicate that case management had a beneficial effect on the clients receiving it. This effect may have been the result of an increase in services received by the case-managed clients.