Instantaneously trained neural networks

Instantaneously trained neural networks are <a href="/facts/Feedforward_neural_network/CP0pPGDF">feedforward artificial neural networks</a> that create a new hidden neuron node for each novel training sample. The weights to this hidden neuron separate out not only this training sample but others that are near it, thus providing generalization. This separation is done using the nearest hyperplane that can be written down instantaneously. In the two most important implementations the neighborhood of generalization either varies with the training sample (CC1 network) or remains constant (CC4 network). These networks use <a href="/facts/Unary_coding/yHF7r7pE">unary coding</a> for an effective representation of the data sets.
This type of network was first proposed in a 1993 paper of <a href="/facts/Subhash_Kak/FQLGwn3n">Subhash Kak</a>. Since then, instantaneously trained neural networks have been proposed as models of short term <a href="/facts/Learning/lehEKRjC">learning</a> and used in <a href="/facts/Web_search/SRntsbz2">web search</a>, and financial <a href="/facts/Time_series_prediction/fSXPR817">time series prediction</a> applications. They have also been used in instant <a href="/facts/Document_classification/B69lL6BH">classification of documents</a> and for <a href="/facts/Deep_learning/JLuwD3ea">deep learning</a> and <a href="/facts/Data_mining/AH53q5ac">data mining</a>.
As in other neural networks, their normal use is as software, but they have also been implemented in hardware using FPGAs and by <a href="/facts/Optical_neural_network/ob66auDd">optical implementation</a>.

Instantaneously trained neural networks open-in-new

Instantaneously trained neural networks