Subscribe to our mailing list
MemComputing
Subscribe

“ We will not share your information.”

Wanna get our awesome news?
We will send you emails only several times per week. Isn't that cool?
Subscribe!

Actually we will not spam you and keep your personal data secure

Try it Free
Try it Free
Try it Free

Latest News

QUANTUM COMPUTING

CARGO LOAD PLANNING

Cargo load planning is computationally intensive and one of those problems that is expected to be solved efficiently by quantum computers. However, quantum computers are still estimated to be 10 years away.

Our latest paper shows how MemComputing can solve these problems today a good decade before quantum computers.

Read Case Study
Read technical publication

CASE STUDY

TRANSPORTATION LOGISTICS

Read our case study that shows how MemComputing can solves NP-Hard port logistics problems in seconds, where today’s best in class solution takes over 70 hours.

Read more

FUNDING

LATEST INVESTMENT

Japanese investment firm, IT-Farm invests in MemComputing.

Read more

What it isn’t

It is not a Turing Machine
It does not follow the Von Neumann Architecture
It isn’t Quantum Computing

What it is

An entirely new paradigm
Computation & Memory Combined in the same circuit
A brand new/patented computer architecture
Uses classical low power, low heat transistor technology
Emulated in software, it runs on classical architecture

Optimization Problems Are Everywhere

These fundamental problems pervade essentially every scientific discipline and industry and are usually difficult to solve

Life Sciences

Drug Design

Finance

Portifolio Optimization

Energy

Demand Forecasting

Machine Learning

Deep Learning

Logistic & Distribution

Resource Planning

Optimization problems involve finding the best solution from all possible solutions; often exploring millions, billions or even more possible combinations
One approach is to exhaustively search and check every possible solution but this “brute force” becomes less and less feasible as problems get bigger

For many optimization problems, the time required to solve grows exponentially as more variables are added often rendering them intractable for modern computers

An Exponential Problem

Exponential problems are often called “intractable” since they can take longer than a lifetime to solve

Currently, many known solutions for optimization problems involve exponential growth
  • Complex optimization problems are often classified as “NP-hard“ or “NP-Complete”.
    “NP” problems are problems which are unsolvable within a reasonable amount of time for large input sizes.
  • The “Time Complexity” of a problem defines how the time to solve increases with input size
  • O(n) – Execution time increases linearly with input size
  • O(n^2 ) – Quadratic Growth
  • O(2^n ) – Exponential Growth

Exponential problems are often called “intractable” since they can take longer than a lifetime to solve

 

What to do?

If most optimization problems are intractable, how do we solve them?
For many years, research has been focused on developing computer systems and techniques that could efficiently solve these NP problems. However many challenges remain.

Select Techniques and Systems for Approaching Intractable Optimization Problems

Branch & Bound

Pros

  • Often can eliminate a large percentage of possible solutions
  • Still finds an optimal solution
  • Usually faster than exhaustive search

  • Still has exponential complexity
  • Still not feasible for problems with a large number of inputs 

Heuristics

Pros

  • Focus on finding a “good” solution (approximation)
  • Research has produced many possible methods / techniques
  • Can be very fast

  • Non-Convex landscapes can produce poor results
  • Sensitive to input parameters
  • Can take a long time to produce near optimal results

Quantum Computing

Pros

  • Theoretically known to speed up optimization problems
  • Takes advantage of physical characteristics to solve problems

  • Extreme system operating environment
  • Scalability is a BIG concern
  • Overcoming system noise is a significant engineering challenge

A technology that can efficiently provide accurate solutions to these exponential problems stands to unlock tremendous value across a number of industrial & scientific verticals

MemComputing’s New Approach

A proprietary computing architecture that achieves an exponential speed-up and finds better solutions

  • Inspired by the computational efficiency of the brain
  • Fundamentally different computational architecture
  • Introduces “Self Organizing Logic Gates”- accepting inputs
 from any terminal which enable collective behavior
  • Takes advantage of non-locality of memory and 
long range correlations between gates
  • Completely avoids the von Neumann Bottleneck

Types of Applications/Customers

MemComputing can scale to address challenging problems for a wide variety of industries and application