IBM Has Achieved Cat Scale Brain Simulation Which are 15 times the scale of Previous Rat Brain Simulations

The Cat is Out of the Bag and BlueMatter

Today at SC 09, the supercomputing conference in Portland, Oregon, IBM is announcing progress toward creating a computer system that simulates the way the brain works. (Brain Emulation) Two major milestones indicate the feasibility of building a cognitive computing chip: unprecedented advances in large-scale cortical simulation and a new algorithm that synthesizes neurological data. This work is on track to human brain scale simulations in 2018.

12 page pdf on IBMs cat brain emulation

IBM simulated a model with 0.9 * 10^9 neurons and 0.9 * 10^13 synapses, using probabilistic connectivity and a simulation time step of 1 ms, only 83 times slower than real-time per Hertz of average neuronal ring rate.

In the quest for cognitive computing, we have built a massively parallel cortical simulator, C2, that incorporates a number of innovations in computation, memory, and communication. Using C2 on LLNL’s Dawn Blue Gene/P supercomputer with 146,456 CPUs and 144 TB of main memory, we report two cortical simulations { at unprecedented scale { that e®ectively saturate the entire memory capacity and refresh it at least every simulated second. The first simulation consists of 1.6 billion neurons and 8.87 trillion synapses with experimentally-measured gray matter thalamocortical connectivity. The second simulation has 900 million neurons and 9 trillion synapses with probabilistic connectivity. We demonstrate nearly perfect weak scaling and attractive strong scaling. The simulations, which incorporate phenomenological spiking neurons, individual learning synapses, axonal delays, and dynamic synaptic channels, exceed the scale of the cat cortex, marking the dawn of a new era in the scale of cortical simulations.

The cat brain has 15 times as many neurons as a rat and 50 times as many as a mouse. The cat has 13 times as many synapses as a rat and and 35 times as many synapses as a mouse. The new simulations are 4.5% of the size of human cortex simulations. They need to be sped up to full speed (they are 83 times slower than real brains.)

Cognition and computation arise from the cerebral cortex; a truly complex system that contains roughly 20 billion neurons and 200 trillion synapses.

Weak Scaling of C2 in each of the three Blue Gene/P modes: SMP (one CPU per node), DUAL (two CPUs per node) and VN (four CPUs per node). The plots show that in each mode, as the number of MPI Processes is increased (x-axis), a proportionately larger size of the model, quanti¯ed with number of synapses, can be successfully simulated (y-axis). Both axes are on a logarithmic scale (with base 2) and the straight line curves have a slope of 1, demonstrating a doubling of model size with doubling of available CPUs. This is nearly perfect weak scaling in memory. The overall model size is enlarged by increasing the number of groups of neurons: for every 1024 nodes added to the simulation, the number of neuron groups was increased by 32,768. This choice of number of neuron groups allows a good degree of load balancing, and also allows the group sizes to be uniform across all the models. The horizontal lines are provided for reference and indicate the number of synapses in the cortex of various mammals of interest (see table in the introduction). On a half-rack system with 512 nodes, we were able to simulate at a scale of the mouse cortex, comparable to our prior work on 2 racks with 2,048 nodes, we were able to simulate at a scale of the rat cortex, comparable to our previous report. Representing previously unattained scales, on 4 racks with 4,096 nodes, we are able to simulate at a scale of the ultimate objective of the SyNAPSE program; with a little over 24,756 nodes and 24 racks, we simulated a 6.1 trillion synapses at the scale of the cat cortex. Finally, the largest model size consists of 900 million neurons and 9 trillion synapses, in 1,179,648 groups of 763 neurons each. This corresponds to a scale of 4.5% of the human cortex.

Using a state-of-the-art Blue Gene/P with 147,456 processors and 144 TB of main memory, we were able to simulate a thalamocortical model at an unprecedented scale of 10^9 neurons and 10^13 synapses. Compared to the human cortex, our simulation has a scale that is roughly 1 to 2 orders smaller and has a speed that is 2 to 3 orders slower than real-time. Our work opens the doors for bottom-up, actual-scale models of the thalamocortical system derived from biologically-measured data. In the very near future, we are planning to further enrich the models with long-distance white-matter connectivity [35]. We are also working to increase the spatial topographic resolution of thalamocortical gray-matter connectivity 100 times { from hypercolumn (» 10,000 neurons) to minicolumn (» 100 neurons).

What does BlueMatter mean?
BlueMatter is a highly parallelized algorithm for identifying white matter projectomes written to take advantage of the Blue Gene supercomputing architecture. Hence, the term BlueMatter.

Can you please provide more details on BlueMatter?
Our software, BlueMatter, is able to provide unique visualization and measurement of the long range circuitry (interior white matter) that allow geographically separated regions of the brain to communicate. The labels or colors of the fibers represent divisions of these fibrous networks that we are measuring. The colors and names are as follows:

Red – Interhemispheric fibers projecting between the corpus callosum and frontal cortex.
Green – Interhemispheric fibers projecting between primary visual cortex and the corpus callosum.
Yellow – Interhemispheric fibers projecting from corpus callosum and not Red or Green.
Brown – Fibers of the superior longitudinal fasciculus, connecting regions critical for language processing.
Orange – Fibers of inferior longitudinal fasciculus and uncinate fasciculus, connecting regions to cortex responsible for memory.
Purple – Projections between parietal lobe and lateral cortex
Blue – Fibers connecting local regions of the frontal cortex

About The Author

Add comment

E-mail is already registered on the site. Please use the Login form or enter another.

You entered an incorrect username or password

Sorry, you must be logged in to post a comment.