Hampshire College

About: Hampshire College is a education organization based out in Amherst Center, Massachusetts, United States. It is known for research contribution in the topics: Genetic programming & Population. The organization has 461 authors who have published 998 publications receiving 40827 citations.

Innovation is built on the obscure: innovation-enhancing software for uncovering the obscure

Abstract: Analysis of over 1,000 innovative inventions reveals that during the innovative process at least one rarely-noticed or new (i.e., obscure) feature is unearthed and built upon to create the solution (i.e., the Obscure Features Hypothesis for innovation: OFH) [6, 7]. Embedding the insights from this analysis into the structure of semantic networks creates AhaNets, which help optimize the search for the needed key obscure feature. Techniques to overcome cognitive aversions to noticing the obscure (i.e., fixation effects) further enhance innovation by improving the search process. Once implemented in software, AhaNets and counter-fixation techniques create an innovation-enhancing human-machine interaction.

Evolution Evolves with Autoconstruction

Abstract: In autoconstructive evolutionary algorithms, individuals implement not only candidate solutions to specified computational problems, but also their own methods for variation of offspring. This makes it possible for the variation methods to themselves evolve, which could, in principle, produce a system with an enhanced capacity for adaptation and superior problem solving power. Prior work on autoconsruction has explored a range of system designs and their evolutionary dynamics, but it has not solved hard problems. Here we describe a new approach that can indeed solve at least some hard problems. We present the key components of this approach, including the use of linear genomes for hierarchically structured programs, a diversity-maintaining parent selection algorithm, and the enforcement of diversification constraints on offspring. We describe a software synthesis benchmark problem that our new approach can solve, and we present visualizations of data from single successful runs of autoconstructive vs. non-autoconstructive systems on this problem. While anecdotal, the data suggests that variation methods, and therefore significant aspects of the evolutionary process, evolve over the course of the autoconstructive runs.

Involving students in experimental design: three approaches.

Abstract: Many faculty want to involve students more actively in laboratories and in experimental design. However, just "turning them loose in the lab" is time-consuming and can be frustrating for both stude...

A Detailed Analysis of a PushGP Run

Abstract: In evolutionary computation we potentially have the ability to save and analyze every detail in an run. This data is often thrown away, however, in favor of focusing on the final outcomes, typically captured and presented in the form of summary statistics and performance plots. Here we use graph database tools to store every parent–child relationship in a single genetic programming run, and examine the key ancestries in detail, tracing back from an solution to see how it was evolved over the course of 20 generations. To visualize this genetic programming run, the ancestry graph is extracted, running from the solution(s) in the final generation up to their ancestors in the initial random population. The key instructions in the solution are also identified, and a genetic ancestry graph is constructed, a subgraph of the ancestry graph containing only those individuals that contributed genetic information (or instructions) to the solution. These visualizations and our ability to trace these key instructions throughout the run allow us to identify general inheritance patterns and key evolutionary moments in this run.