Tag Archives: Neural Networks

Distributing a neural network among the robots of a swarm

Michael Otte, An emergent group mind across a swarm of robots: Collective cognition and distributed sensing via a shared wireless neural network, The International Journal of Robotics Research, DOI: 10.1177/0278364918779704.

We pose the “trained-at-runtime heterogeneous swarm response problem,” in which a swarm of robots must do the following three things: (1) Learn to differentiate between multiple classes of environmental feature patterns (where the feature patterns are distributively sensed across all robots in the swarm). (2) Perform the particular collective behavior that is the appropriate response to the feature pattern that the swarm recognizes in the environment at runtime (where a collective behavior is defined by a mapping of robot actions to robots). (3) The data required for both (1) and (2) is uploaded to the swarm after it has been deployed, i.e., also at runtime (the data required for (1) is the specific environmental feature patterns that the swarm should learn to differentiate between, and the data required for (2) is the mapping from feature classes to swarm behaviors). To solve this problem, we propose a new form of emergent distributed neural network that we call an “artificial group mind.” The group mind transforms a robotic swarm into a single meta-computer that can be programmed at runtime. In particular, the swarm-spanning artificial neural network emerges as each robot maintains a slice of neurons and forms wireless neural connections between its neurons and those on nearby robots. The nearby robots are discovered at runtime. Experiments on real swarms containing up to 316 robots demonstrate that the group mind enables collective decision-making based on distributed sensor data, and solves the trained-at-runtime heterogeneous swarm response problem. The group mind is a new tool that can be used to create more complex emergent swarm behaviors. The group mind also enables swarm behaviors to be a function of global patterns observed across the environment—where the patterns are orders of magnitude larger than the robots themselves.

Using memory of past input data to improve the convergence of NN when faced with small samples

Zhang, S., Huang, K., Zhang, R. et al., Learning from Few Samples with Memory Network, Cogn Comput (2018) 10: 15, DOI: 10.1007/s12559-017-9507-z.

Neural networks (NN) have achieved great successes in pattern recognition and machine learning. However, the success of a NN usually relies on the provision of a sufficiently large number of data samples as training data. When fed with a limited data set, a NN’s performance may be degraded significantly. In this paper, a novel NN structure is proposed called a memory network. It is inspired by the cognitive mechanism of human beings, which can learn effectively, even from limited data. Taking advantage of the memory from previous samples, the new model achieves a remarkable improvement in performance when trained using limited data. The memory network is demonstrated here using the multi-layer perceptron (MLP) as a base model. However, it would be straightforward to extend the idea to other neural networks, e.g., convolutional neural networks (CNN). In this paper, the memory network structure is detailed, the training algorithm is presented, and a series of experiments are conducted to validate the proposed framework. Experimental results show that the proposed model outperforms traditional MLP-based models as well as other competitive algorithms in response to two real benchmark data sets.

Interesting approach to learning the sensorimotor behavior of a robot and of its predictive capabilities through NN

R. Santos, R. Ferreira, Â. Cardoso and A. Bernardino, SNet: Co-Developing Artificial Retinas and Predictive Internal Models for Real Robots, IEEE Transactions on Cognitive and Developmental Systems, vol. 9, no. 3, pp. 213-222, DOI: 10.1109/TCDS.2016.2638885.

This paper focuses on a recently developed biologically inspired architecture, here denoted as sensorimotor network (SNet), able to co-develop sensorimotor structures directly from data acquired by a robot interacting with its environment. Such networks learn efficient internal models of the sensorimotor system, developing simultaneously sensor and motor representations as well as predictive models of the sensorimotor relationships adapted to their operating environment. Here, we describe our recent model of sensorimotor development and compare its performance with neural network models in predicting self-induced stimuli. In addition, we illustrate the influence of available resources and environment characteristics in the development of the SNet structures. Finally, an SNet is trained using real data recorded during a quadricopter drone flight.

Cognitive control: a nice bunch of definitions and state-of-the-art

S. Haykin, M. Fatemi, P. Setoodeh and Y. Xue, Cognitive Control, in Proceedings of the IEEE, vol. 100, no. 12, pp. 3156-3169, Dec. 2012., DOI: 10.1109/JPROC.2012.2215773.

This paper is inspired by how cognitive control manifests itself in the human brain and does so in a remarkable way. It addresses the many facets involved in the control of directed information flow in a dynamic system, culminating in the notion of information gap, defined as the difference between relevant information (useful part of what is extracted from the incoming measurements) and sufficient information representing the information needed for achieving minimal risk. The notion of information gap leads naturally to how cognitive control can itself be defined. Then, another important idea is described, namely the two-state model, in which one is the system’s state and the other is the entropic state that provides an essential metric for quantifying the information gap. The entropic state is computed in the perceptual part (i.e., perceptor) of the dynamic system and sent to the controller directly as feedback information. This feedback information provides the cognitive controller the information needed about the environment and the system to bring reinforcement leaning into play; reinforcement learning (RL), incorporating planning as an integral part, is at the very heart of cognitive control. The stage is now set for a computational experiment, involving cognitive radar wherein the cognitive controller is enabled to control the receiver via the environment. The experiment demonstrates how RL provides the mechanism for improved utilization of computational resources, and yet is able to deliver good performance through the use of planning. The paper finishes with concluding remarks.

Nice summary of reinforcement learning in control (Adaptive Dynamic Programming) and the use of Q-learning plus NN approximators for solving a control problem under a game theory framework

Kyriakos G. Vamvoudakis, Non-zero sum Nash Q-learning for unknown deterministic continuous-time linear systems, Automatica, Volume 61, November 2015, Pages 274-281, ISSN 0005-1098, DOI: 10.1016/j.automatica.2015.08.017.

This work proposes a novel Q-learning algorithm to solve the problem of non-zero sum Nash games of linear time invariant systems with N -players (control inputs) and centralized uncertain/unknown dynamics. We first formulate the Q-function of each player as a parametrization of the state and all other the control inputs or players. An integral reinforcement learning approach is used to develop a model-free structure of N -actors/ N -critics to estimate the parameters of the N -coupled Q-functions online while also guaranteeing closed-loop stability and convergence of the control policies to a Nash equilibrium. A 4th order, simulation example with five players is presented to show the efficacy of the proposed approach.

Detecting objects in images through the timing of the changes in the visual sensor, rather than through the analysis of frames (without time information)

Orchard, G.; Meyer, C.; Etienne-Cummings, R.; Posch, C.; Thakor, N.; Benosman, R., HFirst: A Temporal Approach to Object Recognition, in Pattern Analysis and Machine Intelligence, IEEE Transactions on , vol.37, no.10, pp.2028-2040, Oct. 1 2015 DOI: 10.1109/TPAMI.2015.2392947.

This paper introduces a spiking hierarchical model for object recognition which utilizes the precise timing information inherently present in the output of biologically inspired asynchronous address event representation (AER) vision sensors. The asynchronous nature of these systems frees computation and communication from the rigid predetermined timing enforced by system clocks in conventional systems. Freedom from rigid timing constraints opens the possibility of using true timing to our advantage in computation. We show not only how timing can be used in object recognition, but also how it can in fact simplify computation. Specifically, we rely on a simple temporal-winner-take-all rather than more computationally intensive synchronous operations typically used in biologically inspired neural networks for object recognition. This approach to visual computation represents a major paradigm shift from conventional clocked systems and can find application in other sensory modalities and computational tasks. We showcase effectiveness of the approach by achieving the highest reported accuracy to date (97.5% ± 3.5%) for a previously published four class card pip recognition task and an accuracy of 84.9% ± 1.9% for a new more difficult 36 class character recognition task.

Transfer learning in reinforcement learning through case-based and the use of heuristics for selecting actions

Reinaldo A.C. Bianchi, Luiz A. Celiberto Jr., Paulo E. Santos, Jackson P. Matsuura, Ramon Lopez de Mantaras, Transferring knowledge as heuristics in reinforcement learning: A case-based approach, Artificial Intelligence, Volume 226, September 2015, Pages 102-121, ISSN 0004-3702, DOI: 10.1016/j.artint.2015.05.008.

The goal of this paper is to propose and analyse a transfer learning meta-algorithm that allows the implementation of distinct methods using heuristics to accelerate a Reinforcement Learning procedure in one domain (the target) that are obtained from another (simpler) domain (the source domain). This meta-algorithm works in three stages: first, it uses a Reinforcement Learning step to learn a task on the source domain, storing the knowledge thus obtained in a case base; second, it does an unsupervised mapping of the source-domain actions to the target-domain actions; and, third, the case base obtained in the first stage is used as heuristics to speed up the learning process in the target domain.
A set of empirical evaluations were conducted in two target domains: the 3D mountain car (using a learned case base from a 2D simulation) and stability learning for a humanoid robot in the Robocup 3D Soccer Simulator (that uses knowledge learned from the Acrobot domain). The results attest that our transfer learning algorithm outperforms recent heuristically-accelerated reinforcement learning and transfer learning algorithms.

Reinforcement learning applied to select which parts of a Neural Turing Machine are to be updated with backpropagation during learning

Wojciech Zaremba, Ilya Sutskever, Reinforcement Learning Neural Turing Machines, arXiv.org, arXiv:1505.00521.

The expressive power of a machine learning model is closely related to the number of sequential computational steps it can learn. For example, Deep Neural Networks have been more successful than shallow networks because they can perform a greater number of sequential computational steps (each highly parallel). The Neural Turing Machine (NTM) is a model that can compactly express an even greater number of sequential computational steps, so it is even more powerful than a DNN. Its memory addressing operations are designed to be differentiable; thus the NTM can be trained with backpropagation.
While differentiable memory is relatively easy to implement and train, it necessitates accessing the entire memory content at each computational step. This makes it difficult to implement a fast NTM. In this work, we use the Reinforce algorithm to learn where to access the memory, while using backpropagation to learn what to write to the memory. We call this model the RL-NTM. Reinforce allows our model to access a constant number of memory cells at each computational step, so its implementation can be faster. The RL-NTM is the first model that can, in principle, learn programs of unbounded running time. We successfully trained the RL-NTM to solve a number of algorithmic tasks that are simpler than the ones solvable by the fully differentiable NTM.
As the RL-NTM is a fairly intricate model, we needed a method for verifying the correctness of our implementation. To do so, we developed a simple technique for numerically checking arbitrary implementations of models that use Reinforce, which may be of independent interest.

Mental imaginery for a mobile robot that learns obstacle avoidance

Wilmer Gaona, Esaú Escobar, Jorge Hermosillo, Bruno Lara (2015), Anticipation by multi-modal association through an artificial mental imagery process, Connection Science, 27:1, 68-88, DOI: 10.1080/09540091.2014.95628

Mental imagery has become a central issue in research laboratories seeking to emulate basic cognitive abilities in artificial agents. In this work, we propose a computational model to produce an anticipatory behaviour by means of a multi-modal off-line hebbian association. Unlike the current state of the art, we propose to apply hebbian learning during an internal sensorimotor simulation, emulating a process of mental imagery. We associate visual and tactile stimuli re-enacted by a long-term predictive simulation chain motivated by covert actions. As a result, we obtain a neural network which provides a robot with a mechanism to produce a visually conditioned obstacle avoidance behaviour. We developed our system in a physical Pioneer 3-DX robot and realised two experiments. In the first experiment we test our model on one individual navigating in two different mazes. In the second experiment we assess the robustness of the model by testing in a single environment five individuals trained under different conditions. We believe that our work offers an underpinning mechanism in cognitive robotics for the study of motor control strategies based on internal simulations. These strategies can be seen analogous to the mental imagery process known in humans, opening thus interesting pathways to the construction of upper-level grounded cognitive abilities.