APPARATUSES AND METHODS FOR OPERATING NEURAL NETWORKS
The present disclosure includes apparatuses and methods for operating neural networks. An example apparatus includes a plurality of neural networks, wherein the plurality of neural networks are configured to receive a particular portion of data and wherein each of the plurality of neural networks are configured to operate on the particular portion of data during a particular time period to make a determination regarding a characteristic of the particular portion of data.
1 .- 20 . (canceled)
21 . A processing in memory (PIM) device, comprising:
a number of banks each comprising:
one or more memory arrays; and
a controller coupled to the one or more memory arrays, wherein the controller is configured to operate a neural network on a portion of data representing an image, sound, or emotion, wherein each neural network of the number of banks is configured to operate simultaneously to classify the portion of data.
22 . The device of claim 21 , wherein each neural network is configured to include processing in memory (PIM) architecture.
23 . The device of claim 21 , wherein each neural network includes sensing circuitry including a sense amplifier and a compute component.
24 . The device of claim 21 , wherein each neural network is independently trained.
25 . The device of claim 21 , wherein each neural network is configured to simultaneously receive instructions to operate on the portion of data.
26 . The device of claim 21 , wherein each neural network is configured to operate in a fixed point or binary weighted network.
27 . The device of claim 21 , wherein each neural network is a single-bit network.
28 . The device of claim 21 , wherein the controller comprises control logic, sequencers, and timing circuitry.
29 . A processing in memory (PIM) device, comprising:
a number of banks each comprising:
one or more memory arrays; and
a controller coupled to the one or more memory arrays, wherein the controller is configured to:
operate a neural network on a portion of data, wherein the portion of data represents an image, sound, or emotion and each neural network of the number of banks is configured to operate simultaneously to classify the portion of data; and
weigh an accuracy of data recognition based on results of each neural network.
30 . The device of claim 29 , further comprising a high speed interface (HSI) configured to receive a command and the portion of data from a host.
31 . The device of claim 30 , wherein the HSI is coupled to a bank arbiter.
32 . The device of claim 30 , wherein the controller is configured to operate each neural network on the portion of data in response to receiving the command from the host.
33 . The device of claim 29 , wherein the controller is configured to receive a vote from each neural network.
34 . The device of claim 33 , wherein the vote from each neural network is weighted by the controller.
35 . The device of claim 34 , wherein the vote from each neural network is weighted based on type of particular portion of data and particular training of each neural network.
36 . The device of claim 29 , wherein the controller is configured to weigh the accuracy of the data recognition using a voting scheme.
37 . The device of claim 36 , wherein the voting scheme is a majority rule or an average.
38 . The device of claim 29 , wherein an output is provided by the controller based on the accuracy of the data recognition.
39 . The device of claim 38 , wherein the output is discarded if there is no uniform decision on the accuracy of the data recognition among each neural network.
40 . A method comprising:
operating a neural network on each of a plurality of banks of a processing in memory (PIM) device;
receiving a portion of data representing an image, sound, or emotion at each neural network; and
determining a characteristic of the portion of data simultaneously on each neural network.