Back

A reduced-memory multi-layer perceptron withsystematic network weights generated and trainedthrough distribution hyper-parameters

Ahmad, S.; Vinayak, N.

2022-03-03 bioinformatics
10.1101/2022.03.01.482493 bioRxiv
Show abstract

A multi-layer perceptron (MLP) consists of a number of forward-connected weights (Wijk) from each feeding layer node (nij) to the many initially equivalent nodes (ni+1,k) in the next layer. Exact a priori order and search space of these weights (Wijk) is random and prone to redundancy, irreproducibility and non-optimality. We demonstrate that a weight subspace (Wijk for each i and j), generated systematically using a statistical distribution with predefined breakpoints and Genetic algorithm-trained hyper-parameters substantially reduces the computational complexity of an MLP and produces comparable or better performance than similarly trained equivalent models with fully defined weights. This distribution based neural network (DBNN) provides a novel framework to create very large neural network models with currently prohibitive memory requirements.

Matching journals

The top 10 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.