Efficient and scalable systems for calculating neural network connectivity in an event-driven way
First Claim
1. A scalable system for recalculating, in an event-driven manner, property parameters including connectivity parameters of a neural network, the system comprising:
- an input component that receives a time varying input signal;
a state machine comprising a pseudo-random number generator with a seed value, wherein the state machine is capable of recalculating property parameters of the neural network using the seed value, wherein the property parameters include connectivity among neurons of the neural network, and wherein the connectivity is deterministically generated by the pseudo-random number generator without reference to preexisting connectivity stored in memory, wherein the state machine is capable of generating a unique identifying number for each neuron in the neural network, and, wherein the state machine comprises a neuron identification counter with a first predefined initial value and a neuron connectivity counter with a second predefined initial value, wherein both the first predefined initial value and the second predefined initial value are utilized to update the state machine, and, wherein the pseudo-random number generator of the state machine comprises a Linear Feedback Shift Register (LSFR);
a storage component containing a plurality of predefined seed values from which a particular seed value is chosen for the pseudo-random number generator, wherein the number of seed values is less than the number of property parameters to be recalculated; and
an output component that generates output signals reflective of the calculated property parameters of the neural network and the input signal.
1 Assignment
0 Petitions
Accused Products
Abstract
Systems and methods achieving scalable and efficient connectivity in neural algorithms by re-calculating network connectivity in an event-driven way are disclosed. The disclosed solution eliminates the storing of a massive amount of data relating to connectivity used in traditional methods. In one embodiment, a deterministic LFSR is used to quickly, efficiently, and cheaply re-calculate these connections on the fly. An alternative embodiment caches some or all of the LFSR seed values in memory to avoid sequencing the LFSR through all states needed to compute targets for a particular active neuron. Additionally, connections may be calculated in a way that generates neural networks with connections that are uniformly or normally (Gaussian) distributed.
15 Citations
28 Claims
-
1. A scalable system for recalculating, in an event-driven manner, property parameters including connectivity parameters of a neural network, the system comprising:
-
an input component that receives a time varying input signal; a state machine comprising a pseudo-random number generator with a seed value, wherein the state machine is capable of recalculating property parameters of the neural network using the seed value, wherein the property parameters include connectivity among neurons of the neural network, and wherein the connectivity is deterministically generated by the pseudo-random number generator without reference to preexisting connectivity stored in memory, wherein the state machine is capable of generating a unique identifying number for each neuron in the neural network, and, wherein the state machine comprises a neuron identification counter with a first predefined initial value and a neuron connectivity counter with a second predefined initial value, wherein both the first predefined initial value and the second predefined initial value are utilized to update the state machine, and, wherein the pseudo-random number generator of the state machine comprises a Linear Feedback Shift Register (LSFR); a storage component containing a plurality of predefined seed values from which a particular seed value is chosen for the pseudo-random number generator, wherein the number of seed values is less than the number of property parameters to be recalculated; and an output component that generates output signals reflective of the calculated property parameters of the neural network and the input signal. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15)
-
-
16. A computer-implemented method for recalculating network property parameters of a neural network including connectivity parameters in an event-driven manner, the method comprising:
-
initializing property parameters of the neural network; receiving, at an evaluating neuron of the neural network, a neural input corresponding to a time varying input signal to the neural network; recalculating, by a state machine comprising a pseudo-random number generator with a seed value, at least some of the property parameters of the evaluating neuron, wherein the connectivity parameters are deterministically generated after initialization by the pseudo-random number generator using the seed value without reference to preexisting connectivity stored in memory, wherein the state machine is capable of generating a unique identifying number for each neuron in the neural network, and, wherein the state machine comprises a neuron identification counter with a first predefined initial value and a neuron connectivity counter with a second predefined initial value, wherein both the first predefined initial value and the second predefined initial value are utilized to update the state machine, and wherein the pseudo-random number generator comprises a Linear Feedback Shift Register (LFSR); determining whether the evaluating neuron is to generate a neural output to its target neurons in the neural network; and if the evaluating neuron is determined to generate a neural output to its target neurons in the neural network, propagating the output of the evaluating neuron to its target neurons, wherein a particular seed value is chosen from a plurality of predefined seed values for the pseudo-random number generator, the plurality of predefined seed values being contained in a storage component, and wherein the number of seed values is less than the number of property parameters to be recalculated. - View Dependent Claims (17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28)
-
Specification