Application examples

Neuroblast's neural algorithms are the solution to a wide range of computational problems. Several examples are shown below.

1. Process Control
2. Robotics
3. Coordination and Tactical Planning
4. Modeling and Prediction

Process Control
General-purpose no-tuning automatic control algorithms

The most common process control algorithm in use today is the PID controller. Anyone who has ever tried to use one knows how hard they can be to implement, requiring the simultaneous optimization of two or sometimes three dynamically-coupled tuning parameters. Neuroblast created drop-in alternatives that provide much more power, with much greater ease.

The image on the left shows the disturbance-rejection performance of one of Neuroblast's controllers (in red), compared to an expert-designed PID controller (in grey) on a 20-tray binary distillation column. As usual, the Neuroblast controller was far more robust, returning the process to its setpoint nearly instantly. And yet while the PID controller had to be specially tuned for this specific process, the Neuroblast controller was simple plug and play.

For control problems of greater complexity, Neuroblast can create custom-designed algorithms that are also very simple to use and very robust, making them the ideal solutions for operations in complex, noisy, and unpredictable environments.

Robotics
The coming revolution

The robotics industry is often said to be right now where the PC industry was 30 years ago — on the verge of exploding into a ubiquitous aspect of our everyday lives. But there is one thing holding this industry back: the robots do not have brains.

Neuroblast can create the brains for many different kinds of robotic platforms. For example, the robot on the left, a Pololu 3pi, has five light sensors and two servo motors. Neuroblast developed a neural algorithm to control this robot, enabling it to steer and follow paths at high speed with complete autonomy.

Coordination and Tactical Planning
Adaptive intelligence across time and space

Many different kinds of engineering problems require the efficient and adaptive simultaneous coordination of many moving parts. Examples include air traffic control, street light synchronization, and network packet routing. Neuroblast can create algorithms capabable of cooperating among themselves in the simultaneous pursuit of local and global, short-term and long-term objectives.

To illustrate this capability, Neuroblast has created algorithms to play a game called Alak, a one-dimensional version of the ancient and classic board game Go. The players take turns placing their pieces on the board, attempting to surround and thus capture the pieces of their opponent. The rules are simple, but the tactical and strategic possibilities are boundless.

Neuroblast's neural algorithms play this game with cunning and skill. Anyone wanting to compete their biological brains against Neuroblast's virtual brains may challenge them on our online Alak app.

Modeling and Prediction
Discover hidden trends

Neuroblast's neural algorithms excel at uncovering subtle relationships buried within complex and noisy data streams, and exploiting the predictive value of those relationships.

For example, the image on the left shows a typical trading session of an algorithm that trades simultaneously on two separate securities (in black), and can place orders either long or short (in red; high lines and low lines, respectively). For all the complexity and noise of the underlying process, the algorithm is a very successful trader, posting a typically healthy steadily increasing balance (in green).

Neuroblast is currently developing in partnership with a prominent New York hedge fund an algorithm to discover and analyze ever-shifting trends within a basket of publicly traded stocks, and to recommend buy and sell orders in the pursuit of maximum profit.