What is claimed is:
1. A one-out-of-N neural network encoder and decoder system comprising:
- (a) a set of N input terminals for designating one-out-of-N input states by applying a signal level to one terminal of the set of N input terminals;
- (b) encoding means connected to the set of N input terminals for producing, at an output terminal, a single multilevel output signal having N possible unique levels, each unique level corresponding to one of the designated one-out-of-N input states; and
- (c) docoding means with an input terminal coupled to the encoder means output terminal for receiving a signal representative of the level of the encoding means single output signal, the decoding means having N output terminals, each of the output terminals corresponding to one of the set of N input terminals connected to the encoding means, for decoding the received signal by generating, at the N output terminals, N signals such that the output terminal that corresponds to the designated input state has a value greater than any of the other N-1 output signal level values.
2. A one-out-of-N neural network encoder and decoder system comprising:
- (a) a set of N input terminals for designating one-out-of-N input states by applying a signal level to one terminal of the set of N input terminals:
- (b) encoding means comprising an N-input artificial neuron,, the artificial neuron N-inputs connected to the set of N input terminals for producing at the artificial neuron output terminal, a single multilevel output signal having N possible unique levels, each unique level corresponding to one of the designated one-out-of-N input states; and
- (c) decoding means comprising N single-input artificial neurons with each of the N single-inputs coupled to the encoding means output terminal for receiving a signal representative of the encoding means output signal level, each of the N single-input neurons having an offset input terminal and an output terminal, each of the output terminals corresponding to one of the set of N input terminals connected to the encoding means for decoding the received signal by generating, at the N output terminals, N signal levels such that the output terminal that corresponds to the designated input state has a value greater than any of the other N-1 output signal level values.
3. The system of claim 2 wherein:
- (a) the N-input artificial neuron of the encoding means has exponentially scaled uniform increment synaptic weights;
- (b) the N single-input artificial neurons of the decoding means have linearly scaled uniform increment synaptic weights; and
- (c) the offset input terminals of the N single-input artificial neurons have exponentially scaled uniform increment offset values applied.
4. The system of claim 2 wherein:
- (a) the N-input artificial neuron of the encoding means has linearly scaled uniform increment synaptic weights;
- (b) the N single-input artificial neurons of the decoding means have linearly scaled uniform increment synaptic weights; and
- (c) the offset input terminals of the N single-input artificial neurons have quadratically scaled uniform increment offset values applied.
5. The system of claim 2 wherein:
- (a) the N-input artificial neuron of the encoding means has linearly scaled uniform increment synaptic weights;
- (b) the N single-input artificial neurons of the decoding means have logarithmically scaled uniform increment synaptic weights; and
- (c) the offset input terminals of the N single-input artificial neurons have linearly scaled uniform increment offset values applied.
6. The system of claim 2 wherein:
- (a) The N-input artificial neuron has a set of N synaptic weights (v.sub.1, v.sub.2, . . . , v.sub.j, . . . , v.sub.N), one for each of the N input states;
- (b) the k.sup.th neuron of the N single-input artificial neurons has a synaptic weight value, w.sub.k, and an offset value, .theta..sub.k, where k=1, 2, . . . , N, for producing at the output of the k.sup.th single-input artificial neuron a signal value z.sub.k, proportional to the value given by
- z.sub.k =v.sub.j w.sub.k + .theta..sub.k
- such that a maximum decoder output value occurs at the decoder output terminal k that corresponds to the designated input terminal j, (k=j).
7. The system of claim 3, further comprising an output classifier means connected to the N output terminals of the decoding means, the output classifier means having N output terminals, only one output terminal being active at a time, the active output terminal being selected to correspond to the decoding means output terminal having the greatest output value.
8. The system of claim 4, further comprising an output classifier means connected to the N output terminals of the decoding means, the output classifier means having N output terminals, only one output terminal being active at a time, the active output terminal being selected to correspond to the decoding means output terminal having the greatest output value.
9. The system of claim 5, further comprising an output classifier means connected to the N output terminals of the decoding means, the output classifier means having N output terminals, only one output terminal being active at a time, the active output terminal being selected to correspond to the decoding means output terminal having the greatest output value.
10. The system of claim 7 wherein the output classifier means is a feedforward Hamming-net maximum-likelihood classifier.
11. The system of claim 8 wherein the output classifier means is a feedforward Hamming-net maximum-likelihood classifier.
12. The system of claim 9 wherein the output classifier means is a feedforward Hamming-net maximum-likelihood classifier.
