Neural network
View Patent ↗A neural network includes first electrode lines in parallel, second electrode lines in parallel, a ferroelectric layer, neuron circuits, a first direction control circuit, and a second direction control circuit. The second electrode lines extend in a direction different from the first electrode lines. The ferroelectric layer is arranged between the first electrode lines and the second electrode lines. The neuron circuits are provided in the first electrode lines, respectively. The first direction control circuit is connected between the neuron circuits and the first electrode lines. The second direction control circuit is connected between the neuron circuits and the second electrode lines. The first electrode lines and the second electrode lines are capacitively coupled to form synapse devices at intersections in a plan view, each of the intersections being a portion where a first electrode line and a second electrode line intersect with each other.
1 . A neural network comprising:
first electrode lines in parallel;
second electrode lines that extend in a direction different from the first electrode lines;
a ferroelectric layer arranged between the first electrode lines and the second electrode lines;
neuron circuits provided in the first electrode lines, respectively;
a first direction control circuit connected between the neuron circuits and the first electrode lines; and
a second direction control circuit connected between the neuron circuits and the second electrode lines, wherein
the first electrode lines and the second electrode lines are capacitively coupled to form synapse devices at intersections in a plan view, each of the intersections being a portion where a first electrode line and a second electrode line intersect with each other.
2 . The neural network according to claim 1 , being capable of training a provided signal and retrieving a trained signal, wherein
each of the neuron circuits is configured to provide a signal in a first state or a second state in accordance with an input, and
in training, the first direction control circuit and the second direction control circuit set capacitances of the synapse devices by applying to the first electrode lines and the second electrode lines, voltages different in accordance with states of output signals from the neuron circuits.
3 . The neural network according to claim 2 , wherein
the ferroelectric layer contains bismuth lanthanum titanate, lead zirconate titanate, or barium titanate.
4 . The neural network according to claim 2 , wherein
in retrieval,
the first direction control circuit applies a voltage lower than in training to the first electrode lines in accordance with an input signal, and
the second electrode lines provide voltages weighted in accordance with the capacitances of the synapse devices set in training.
5 . The neural network according to claim 4 , wherein
the ferroelectric layer contains bismuth lanthanum titanate, lead zirconate titanate, or barium titanate.
6 . The neural network according to claim 4 , wherein
in training, the first direction control circuit and the second direction control circuit
apply, when outputs from the neuron circuits are in the first state, a voltage to the first electrode lines and the second electrode lines so as to apply a first voltage to the synapse devices, and
apply, when outputs from the neuron circuits are in the second state, a voltage to the first electrode lines and the second electrode lines so as to apply a second voltage lower than the first voltage to the synapse devices, and
in the synapse devices, a capacitance when the first voltage is applied is different from a capacitance when the second voltage is applied.
7 . The neural network according to claim 6 , wherein
in retrieval, the first direction control circuit applies a third voltage lower than the first voltage and the second voltage to the first electrode lines.
8 . The neural network according to claim 6 , wherein
the ferroelectric layer contains bismuth lanthanum titanate, lead zirconate titanate, or barium titanate.
9 . The neural network according to claim 7 , wherein
the third voltage has a voltage value at a level at which the capacitances of the synapse devices are not varied.
10 . The neural network according to claim 7 , wherein
the ferroelectric layer contains bismuth lanthanum titanate, lead zirconate titanate, or barium titanate.
11 . The neural network according to claim 9 , wherein
the ferroelectric layer contains bismuth lanthanum titanate, lead zirconate titanate, or barium titanate.
12 . The neural network according to claim 1 , wherein
each of the first electrode lines is formed of platinum or silicide of platinum, and
each of the second electrode lines is formed of gold or silicide of gold.
13 . The neural network according to claim 12 , wherein
the ferroelectric layer contains bismuth lanthanum titanate, lead zirconate titanate, or barium titanate.
14 . The neural network according to claim 1 , wherein
each of the first electrode lines is formed of gold or silicide of gold, and
each of the second electrode lines is formed of platinum or silicide of platinum.
15 . The neural network according to claim 1 , wherein
the ferroelectric layer contains bismuth lanthanum titanate, lead zirconate titanate, or barium titanate.
16 . The neural network according to claim 14 , wherein
the ferroelectric layer contains bismuth lanthanum titanate, lead zirconate titanate, or barium titanate.