Jonathan Edwards

Jonathan Edwards, Tomas Petricek

Many beloved programming systems are image-based: self-contained worlds that persist both code and data in a single file. Examples include Smalltalk, LISP, HyperCard, Flash, and spreadsheets. Image-based programming avoids much of the complexity of modern programming technology stacks and encourages more casual and exploratory programming. However conventional file-based programming has better support for collaboration and deployment. These problems have been blamed for the limited commercial success of Smalltalk. We propose to enable collaboration in image-based programming via types and structure editing. We focus on the problem of schema change on persistent data. We turn to static types, which paradoxically require more schema change but also provide a mechanism to express and execute those changes. To determine those changes we turn to structure editing, so that we can capture changes in type definitions with sufficient fidelity to automatically adapt the data to suit. We conjecture that typical schema changes can be handled through structure editing of static types. That positions us to tackle collaboration with what could be called version control for structure editing. We present a theory realizing this idea, which is our main technical contribution. While we focus here on editing types, if we can extend the approach to cover the entire programming experience then it would offer a new way to collaborate in image-based programming.

Saeed Rahimi Gorji, Ole-Christoffer Granmo, Sondre Glimsdal et al.

The Tsetlin Machine (TM) is a machine learning algorithm founded on the classical Tsetlin Automaton (TA) and game theory. It further leverages frequent pattern mining and resource allocation principles to extract common patterns in the data, rather than relying on minimizing output error, which is prone to overfitting. Unlike the intertwined nature of pattern representation in neural networks, a TM decomposes problems into self-contained patterns, represented as conjunctive clauses. The clause outputs, in turn, are combined into a classification decision through summation and thresholding, akin to a logistic regression function, however, with binary weights and a unit step output function. In this paper, we exploit this hierarchical structure by introducing a novel algorithm that avoids evaluating the clauses exhaustively. Instead we use a simple look-up table that indexes the clauses on the features that falsify them. In this manner, we can quickly evaluate a large number of clauses through falsification, simply by iterating through the features and using the look-up table to eliminate those clauses that are falsified. The look-up table is further structured so that it facilitates constant time updating, thus supporting use also during learning. We report up to 15 times faster classification and three times faster learning on MNIST and Fashion-MNIST image classification, and IMDb sentiment analysis.