
The Integration of Connectionism and FirstOrder Knowledge Representation and Reasoning as a Challenge for Artificial Intelligence
Intelligent systems based on firstorder logic on the one hand, and on a...
read it

Applying Deep Learning to Derivatives Valuation
The universal approximation theorem of artificial neural networks states...
read it

Hierarchical Reinforcement Learning with Abductive Planning
One of the key challenges in applying reinforcement learning to reallif...
read it

Generalisation in Neural Networks Does not Require Feature Overlap
That shared features between train and test data are required for genera...
read it

Learning and TNorms Theory
Deep learning has been shown to achieve impressive results in several do...
read it

Enforcing Constraints on Outputs with Unconstrained Inference
Increasingly, practitioners apply neural networks to complex problems in...
read it

Compositional Processing Emerges in Neural Networks Solving Math Problems
A longstanding question in cognitive science concerns the learning mecha...
read it
Artificial Neural Networks that Learn to Satisfy Logic Constraints
Logicbased problems such as planning, theorem proving, or puzzles, typically involve combinatoric search and structured knowledge representation. Artificial neural networks are very successful statistical learners, however, for many years, they have been criticized for their weaknesses in representing and in processing complex structured knowledge which is crucial for combinatoric search and symbol manipulation. Two neural architectures are presented, which can encode structured relational knowledge in neural activation, and store bounded First Order Logic constraints in connection weights. Both architectures learn to search for a solution that satisfies the constraints. Learning is done by unsupervised practicing on problem instances from the same domain, in a way that improves the networksolving speed. No teacher exists to provide answers for the problem instances of the training and test sets. However, the domain constraints are provided as prior knowledge to a loss function that measures the degree of constraint violations. Iterations of activation calculation and learning are executed until a solution that maximally satisfies the constraints emerges on the output units. As a test case, blockworld planning problems are used to train networks that learn to plan in that domain, but the techniques proposed could be used more generally as in integrating prior symbolic knowledge with statistical learning
READ FULL TEXT
Comments
There are no comments yet.