Differential non-volatile memory cell for artificial neural network

• US 10,643,119 B2
• Filed: 05/07/2019
• Issued: 05/05/2020
• Est. Priority Date: 07/24/2018
• Status: Active Grant

• Current Assignee: SanDisk Technologies LLC (Western Digital Corporation)
• Original Assignee: SanDisk Technologies LLC (Western Digital Corporation)
• Inventors: Chiu, Pi-Feng, Choi, Won Ho, Ma, Wen, Lueker-Boden, Martin
• Primary Examiner(s): Hidalgo, Fernando
• Application Number:
  US16/405,178
• Publication Number:
  US 20200034686A1
• Time in Patent Office: 364 Days
• Field of Search: 365148
• CPC Class Codes: G06N 3/04 Architecture, e.g. interconnection topology
• CPC Class Codes: G06N 3/063 using electronic means
• CPC Class Codes: G06N 3/065 Analogue means
• CPC Class Codes: G06N 3/08 Learning methods
• CPC Class Codes: G11C 11/1657 Word-line or row circuits
• CPC Class Codes: G11C 11/1673 Reading or sensing circuits or methods
• CPC Class Codes: G11C 11/54 using elements simulating biological cells, e.g. neuron
• CPC Class Codes: G11C 11/5607 using magnetic storage elements
• CPC Class Codes: G11C 11/5678 using amorphous/crystalline phase transition storage elements
• CPC Class Codes: G11C 11/5685 using storage elements comprising metal oxide memory material, e.g. perovskites
• CPC Class Codes: G11C 13/0004 comprising amorphous/crystalline phase transition cells
• CPC Class Codes: G11C 13/0028 Word-line or row circuits
• CPC Class Codes: G11C 13/003 Cell access
• CPC Class Codes: G11C 13/004 Reading or sensing circuits or methods
• CPC Class Codes: G11C 16/0483 comprising cells having several storage transistors connected in series
• CPC Class Codes: G11C 16/26 Sensing or reading circuits; Data output circuits
• CPC Class Codes: G11C 2213/79 Array wherein the access device being a transistor

Abstract

Use of a non-volatile memory array architecture to realize a neural network (BNN) allows for matrix multiplication and accumulation to be performed within the memory array. A unit synapse for storing a weight of a neural network is formed by a differential memory cell of two individual memory cells, such as a memory cells having a programmable resistance, each connected between a corresponding one of a word line pair and a shared bit line. An input is applied as a pattern of voltage values on word line pairs connected to the unit synapses to perform the multiplication of the input with the weight by determining a voltage level on the shared bit line. The results of such multiplications are determined by a sense amplifier, with the results accumulated by a summation circuit.

Claims

1. An apparatus, comprising:

an array of non-volatile memory cells, including a first bit line, a plurality of word line pairs and a plurality of non-volatile memory cell pairs, each word line of a word line pair being connected to the first bit line by one of the memory cells of a corresponding memory cell pair;

one or more control circuits connected to the array of non-volatile memory cells, the one or more control circuits configured to;

receive one or more binary inputs;

apply one or more voltage patterns, each corresponding to one of the binary inputs, to a corresponding one or more selected word line pairs of the plurality of word line pairs, a first value of the binary input corresponding to a high voltage level on a first of the selected word line pair and a low voltage level a second of the word line pair and a second value of the binary input corresponding to the low voltage level on the first of the selected word line pair and the high voltage level the second of the word line pair; and

determine a voltage level on the first bit line in response to applying the one or more voltage patterns the corresponding one or more selected word line pairs.

2. The apparatus of claim 1, wherein the one or more binary inputs are a plurality of binary inputs and the one or more control circuits are further configured to:

apply the corresponding plurality of voltage patterns sequentially;

determine the voltage level on the first bit line in response to applying the voltage patterns to the corresponding selected word line pairs sequentially; and

accumulate a result of the sequential determination of voltage level on the first bit line in response to applying the voltage patterns the corresponding selected word line pairs.

3. The apparatus of claim 1, wherein the one or more binary inputs are a plurality of binary inputs and the one or more control circuits are further configured to:

apply multiple ones of the corresponding plurality of voltage patterns concurrently; and

determine the voltage level on the first bit line in response to applying the voltage patterns to the corresponding selected word line pairs concurrently as a multi-bit value.

4. The apparatus of claim 1, wherein the one or more control circuits are further configured to:

concurrently with applying the voltage pattern corresponding to the one or more voltage patterns to the one or more selected word line pairs, set both word lines of non-selected word line pairs to float.

5. The apparatus of claim 1, wherein the non-volatile memory cells store data in a binary format.

6. The apparatus of claim 1, wherein the non-volatile memory cells store data in multi-level format.

7. The apparatus of claim 1, wherein the memory cells are part of a monolithic three-dimensional semiconductor memory device where the memory cells are arranged in multiple physical levels above a substrate.

8. The apparatus of claim 7, wherein the array of non-volatile memory cells is formed according to a cross-point architecture.

9. The apparatus of claim 1, wherein the non-volatile memory cells include a transistor connected in series with a programmable resistance.

10. The apparatus of claim 1, wherein the non-volatile memory cells include a switch connected in series with a programmable resistance.

11. The apparatus of claim 1, wherein the non-volatile memory cells include phase change memory (PCM) based memory cells.

12. The apparatus of claim 1, wherein the non-volatile memory cells include phase resistive random access memory (ReRAM) based memory cells.

13. The apparatus of claim 1, wherein the non-volatile memory cells include phase magnetoresistive memory (MRAM) based memory cells.

14. A method, comprising:

receiving a plurality of input values;

translating each of the plurality of input values into a corresponding first voltage pattern, each first voltage pattern being one of a plurality of voltage patterns comprising a pair of voltage values;

applying the plurality of first voltage patterns to one or more pairs of word lines each connected through a first corresponding pair of non-volatile memory cells to a first shared bit line, wherein no more than one first voltage pattern is applied to any single pair of word lines at a time; and

determining one or more voltage levels on the first shared bit line in response to applying the plurality of first voltage patterns to the one or more pairs of word lines.

15. The method of claim 14, wherein the one or more pairs of word lines each further connected through a second corresponding pair of non-volatile memory cells to a second shared bit line and the method further includes:

determining one or more voltage levels on the second shared bit line in response to applying the plurality of first voltage patterns to the one or more pairs of word lines concurrently with determining the one or more voltage levels on the first shared bit line in response to applying the plurality of first voltage patterns to the one or more pairs of word lines.

16. The method of claim 14, wherein:

applying the plurality of first voltage patterns to one or more pairs of word lines connected through the first corresponding pair of non-volatile memory cells to the first shared bit line includes applying the plurality of first voltage patterns sequentially to a corresponding plurality of the word line pairs connected through the first corresponding pair of non-volatile memory cells to the first shared bit line; and

determining the voltage levels on the first shared bit line in response to applying the plurality of first voltage patterns to the plurality of pairs of word lines sequentially.

17. The method of claim 14, wherein:

applying the plurality of first voltage patterns to one or more pairs of word lines connected through the first corresponding pair of non-volatile memory cells to the first shared bit line includes applying the plurality of first voltage patterns concurrently to a corresponding plurality of the word line pairs connected through the first corresponding pair of non-volatile memory cells to the first shared bit line; and

determining voltage level on the first shared bit line in response to applying the plurality of first voltage patterns concurrently to the plurality of pairs of word lines includes performing a multi-bit sensing operation.

18. A non-volatile memory circuit, comprising:

an array of a plurality of non-volatile memory cells connected between one of a plurality of word lines and one of a plurality of bit lines, the non-volatile memory cells configured to store weights of a neural network with each weight stored in a pair of the non-volatile memory cells, each of the memory cells of each pair connected between a distinct one of a pair of word lines and a shared bit line; and

one or more control circuits connected to the array of non-volatile memory cells, the one or more control circuits configured to;

receive one or more inputs for a layer of a neural network and convert the one or more inputs into a corresponding one of a set of voltage patterns;

apply the one or more voltage patterns to word lines of the array of non-volatile memory cells to thereby perform an in-array multiplication of the one or more inputs with the weights; and

accumulate results of the in-array multiplication.

19. The non-volatile memory circuit of claim 18, wherein the one or more inputs for the layer of the neural network are a plurality of inputs for the layer of the neural network and the one or more control circuits are configured to apply the corresponding plurality of voltage patterns to word lines of the array of non-volatile memory cells sequentially.

20. The non-volatile memory circuit of claim 18, wherein the one or more inputs for the layer of the neural network are a plurality of inputs for the layer of the neural network and the one or more control circuits are configured to apply multiple ones of the corresponding plurality of voltage patterns to word lines of the array of non-volatile memory cells concurrently.