Artificial neural networks with precision weight for artificial intelligence
Methods, systems, and devices for an artificial neural network are described. In one example, an artificial neuron in an artificial neural network may include a resistor coupled with an input line and configured to indicate a synaptic weight and a fuse coupled with the resistor. The artificial neuron may also include a selection component coupled with the fuse and configured to activate the fuse for programming the resistor, and a second selection component coupled with the resistor and an output line, the second selection component configured to select the resistor for a read operation.
1. A device for an artificial neural network, comprising:
an input line;
an output line; and
an artificial neuron comprising:
a resistor having a first terminal coupled with the input line to indicate a synaptic weight;
a fuse coupled with the resistor at a second terminal of the resistor;
a first selection component having a first terminal and a second terminal, the second terminal of the first selection component coupled to a voltage source, and the first terminal of the first selection component coupled with the fuse and the resistor at the second terminal of the resistor to activate the fuse for programming the resistor to adjust the synaptic weight based at least in part on varying a resistance of the resistor, the fuse operable to connect and disconnect the resistor from the artificial neuron based at least in part on whether the first selection component activates the fuse; and
a second selection component coupled with the resistor and the output line to select the resistor for a read operation to determine the adjusted synaptic weight.
2. The device of claim 1 , further comprising:
the voltage source coupled with the artificial neuron via the first selection component, and configured to activate the fuse for programming the resistor, or select the resistor for the read operation, or both.
3. The device of claim 2 , wherein the voltage source is further configured to:
apply a voltage signal to the first selection component to activate the fuse for programming the resistor.
4. The device of claim 2 , wherein the first selection component coupled with the fuse is virtually grounded.
5. The device of claim 2 , wherein the voltage source is further configured to:
apply a voltage signal to the second selection component to couple the resistor and the input line with the output line.
6. The device of claim 5 , further comprising:
a controller coupled with the artificial neuron to read a value of the resistor configured to indicate the synaptic weight of the artificial neuron based at least in part on coupling the resistor and the input line with the output line.
7. The device of claim 2 , wherein the voltage source is coupled with a third selection component that is coupled with the fuse and the resistor, wherein selecting the resistor for the read operation is further based at least in part on coupling the third selection component with the fuse and the resistor.
8. The device of claim 2 , wherein the artificial neuron further comprises:
a second resistor coupled with the input line to indicate the synaptic weight;
a second fuse coupled with the second resistor;
a fourth selection component coupled with the second fuse to activate the second fuse for programming the second resistor; and
a fifth selection component coupled with the second resistor and the output line to select the second resistor for the read operation.
9. The device of claim 8 , wherein the voltage source is further configured to:
apply a voltage signal to the fourth selection component to decouple the second fuse from the second resistor based at least in part on a programming operation.
10. The device of claim 9 , further comprising:
a controller coupled with the artificial neuron to disconnect the second resistor from the artificial neuron based at least in part on decoupling the second fuse from the second resistor.
11. The device of claim 8 , wherein the adjusted synaptic weight of the artificial neuron is based at least in part on the resistance of the resistor and a resistance of the second resistor.
12. The device of claim 1 , wherein the resistor comprises a precision metal resistor.
13. An artificial neuron of an artificial neural network, comprising:
a resistor having a first terminal coupled with an input line to indicate a synaptic weight;
a fuse coupled with the resistor at a second terminal of the resistor;
a first selection component having a first terminal and a second terminal, the second terminal coupled to a voltage source, and the first terminal coupled with the fuse and the resistor at the second terminal of the resistor to activate the fuse for programming the resistor to adjust the synaptic weight based at least in part on varying a resistance of the resistor, the fuse operable to connect and disconnect the resistor from the artificial neuron based at least in part on whether the first selection component activates the fuse; and
a second selection component coupled with the resistor and the output line to select the resistor for a read operation to determine the adjusted synaptic weight.