Displaying 1 to 5 from 5 results

opencog - A framework for integrated Artificial Intelligence & Artificial General Intelligence (AGI)

  •    Scheme

OpenCog is a framework for developing AI systems, especially appropriate for integrative multi-algorithm systems, and artificial general intelligence systems. Though much work remains to be done, it currently contains a functional core framework, and a number of cognitive agents at varying levels of completion, some already displaying interesting and useful functionalities alone and in combination. With the exception of MOSES and the CogServer, all of the above are in active development, are half-baked, poorly documented, mis-designed, subject to experimentation, and generally in need of love an attention. This is where experimentation and integration are taking place, and, like any laboratory, things are a bit fluid and chaotic.

tictactoe - Tic Tac Toe Machine Learning

  •    Javascript

Making computers learn how to play tic-tac-toe. I started messing around with reinforcement learning when I heard about a Flappy Bird RL project on GitHub. Out of the box, the algorithm took about 7 or 8 hours to train. I figured it could learn faster if multiple instances of the same algorithm spread out over the internet could all work to update the same matrix. So after forking the repo and creating a distributed learning version, I was able to get it to train in about 30 minutes with 8 browser tabs. I found myself wanting to explore reinforcement learning a bit more, and so this little project was born.

nonce2vec - This is the repo accompanying the paper "High-risk learning: acquiring new word vectors from tiny data" (Herbelot & Baroni, 2017)

  •    Python

A. Herbelot and M. Baroni. 2017. High-risk learning: Acquiring new word vectors from tiny data. Proceedings of EMNLP 2017 (Conference on Empirical Methods in Natural Language Processing). Distributional semantics models are known to struggle with small data. It is generally accepted that in order to learn 'a good vector' for a word, a model must have sufficient examples of its usage. This contradicts the fact that humans can guess the meaning of a word from a few occurrences only. In this paper, we show that a neural language model such as Word2Vec only necessitates minor modifications to its standard architecture to learn new terms from tiny data, using background knowledge from a previously learnt semantic space. We test our model on word definitions and on a nonce task involving 2-6 sentences' worth of context, showing a large increase in performance over state-of-the-art models on the definitional task.

timbl - TiMBL implements several memory-based learning algorithms.

  •    C++

TiMBL is an open source software package implementing several memory-based learning algorithms, among which IB1-IG, an implementation of k-nearest neighbor classification with feature weighting suitable for symbolic feature spaces, and IGTree, a decision-tree approximation of IB1-IG. All implemented algorithms have in common that they store some representation of the training set explicitly in memory. During testing, new cases are classified by extrapolation from the most similar stored cases. For over fifteen years TiMBL has been mostly used in natural language processing as a machine learning classifier component, but its use extends to virtually any supervised machine learning domain. Due to its particular decision-tree-based implementation, TiMBL is in many cases far more efficient in classification than a standard k-nearest neighbor algorithm would be.