In the present article, we will introduce new mechanics of interconnection between neuron firing rate homeostasis and weight change by means of STDP growth bounded by abstract protein reserve, controlled by the intracellular optimization algorithm. We will show, how these cellular dynamics help neurons to filter out the intense signals to help neurons keep a stable firing rate. We will also examine that such filtering does not affect the ability of neurons to recognize the correlated inputs in unsupervised mode. Such an approach might be used in the machine learning domain to improve the robustness of AI systems.
In the present paper, we discuss such similarities among biological and technical systems. We propose the addition to the well-known STDP synaptic plasticity rule to directs the weight modification towards the state associated with the maximal difference between the background noise and correlated signals. The principle of physically constrained weight growth is used as a basis for such control of the modification of the weights. It is proposed, that biological synaptic straight modification is restricted by the existence and production of bio-chemical ‘substances’ needed for plasticity development. In this paper, the information about the noise-to-signal ratio is used to control such a substances’ production and storage and to drive the neuron’s synaptic pressures towards the state with the best signal-to-noise ratio. Several experiments with different input signal regimes are considered to understand the functioning of the proposed approach.
Nikitin, O., Lukyanova, O. Homeostatic Neural Network for Adaptive Control: Examination and Comparison. Lecture Notes in Computer Science (LNCS), vol. 10994, pp. 223-235. Springer, 2018. doi:10.1007/978-3-319-97628-0
Functioning of the biologically inspired neural network with cellular homeostasis is studied in the paper. The network is applied to the task of the control of agent behavior in the stochastic multi-goal environment. Importance of different aspects of the approach is studied on the setups with partially disabled features of the model. It is shown that only full model, incorporating both cellular homeostasis and homeostatically dependent weight correction rule led to the emergence of adaptive behavior. The proposed model is also compared to the Q(λ) reinforcement learning on the same task with multiple goals. Results, illustrating the comparison between Q(λ) and homeostatic neural network, show that proposed approach outperforms conventional in terms of adaptivity, quality of control and convergence speed.
In this article, we propose the approach to procedural optimization of a neural network, based on the combination of information theory and braid theory. The network studied in the article implemented with the intersections between the braid strands, as well as simplified networks (a network with strands without intersections and a simple convolutional deep neural network), are used to solve various problems of multiclass image classification that allow us to analyze the comparative effectiveness of the proposed architecture. The simulation results showed BraidNet’s comparative advantage in learning speed and classification accuracy.
Nikitin, O. and Lukyanova, O. Protozoa behavior reproduced by the combination of genetic optimization and learning. Procedia Computer Science, 2019, Vol. 150C, pp. 179-185. https://doi.org/10.1016/j.procs.2019.02.035
The article is dedicated to the reproduction of behavior of a class of protozoans (ciliates) using computer simulation and genetic optimization. Protozoans have minimal cognitive abilities and no neural system. The only basis of their control system is mRNA and protein systems inside the cell. Ciliates have primitive sensitive receptors which can percept information only from the closest environment. But, nevertheless, unicellular organisms use their cognitive capacities for obstacles avoiding. In this paper, ethology of ciliate Uronychia transfuga is reproduced using a simple GasNet neural network and genetic optimization. It is showed that neural network of only 7 neurons combined with simple learning model is enough to achieve robust and natural behavior.
Lukyanova, O., Nikitin, O. Isotopic Inheritance: A Topological Approach to Genotype Transfer. Lecture Notes in Computer Science (LNCS), vol. 10994, pp. 27-38. Springer, 2018. doi:10.1007/978-3-319-97628-0
We propose a mathematical definition of the abstract notion of genetic variability and a new approach to the problem of the genotype transition to a new generation called “isotopic inheritance”. To this extent, we apply the notion of topological isotopy and use the branch of low-dimensional topology, known as braid theory to encode the values of a gene of arbitrary length into the genome and to propagate the isomorphic genotypes among the offspring. We illustrate the propositional variations of such encoding in different applications. Multiple knapsack problem was solved using the proposed approach.