-
Programming molecular systems to emulate a learning spiking neuron
Authors:
Jakub Fil,
Neil Dalchau,
Dominique Chu
Abstract:
Hebbian theory seeks to explain how the neurons in the brain adapt to stimuli, to enable learning. An interesting feature of Hebbian learning is that it is an unsupervised method and as such, does not require feedback, making it suitable in contexts where systems have to learn autonomously. This paper explores how molecular systems can be designed to show such proto-intelligent behaviours, and pro…
▽ More
Hebbian theory seeks to explain how the neurons in the brain adapt to stimuli, to enable learning. An interesting feature of Hebbian learning is that it is an unsupervised method and as such, does not require feedback, making it suitable in contexts where systems have to learn autonomously. This paper explores how molecular systems can be designed to show such proto-intelligent behaviours, and proposes the first chemical reaction network (CRN) that can exhibit autonomous Hebbian learning across arbitrarily many input channels. The system emulates a spiking neuron, and we demonstrate that it can learn statistical biases of incoming inputs. The basic CRN is a minimal, thermodynamically plausible set of micro-reversible chemical equations that can be analysed with respect to their energy requirements. However, to explore how such chemical systems might be engineered de novo, we also propose an extended version based on enzyme-driven compartmentalised reactions. Finally, we also show how a purely DNA system, built upon the paradigm of DNA strand displacement, can realise neuronal dynamics. Our analysis provides a compelling blueprint for exploring autonomous learning in biological settings, bringing us closer to realising real synthetic biological intelligence.
△ Less
Submitted 9 May, 2022;
originally announced June 2022.
-
A thermodynamically consistent chemical spiking neuron capable of autonomous Hebbian learning
Authors:
Jakub Fil,
Dominique Chu
Abstract:
We propose a fully autonomous, thermodynamically consistent set of chemical reactions that implements a spiking neuron. This chemical neuron is able to learn input patterns in a Hebbian fashion. The system is scalable to arbitrarily many input channels. We demonstrate its performance in learning frequency biases in the input as well as correlations between different input channels. Efficient compu…
▽ More
We propose a fully autonomous, thermodynamically consistent set of chemical reactions that implements a spiking neuron. This chemical neuron is able to learn input patterns in a Hebbian fashion. The system is scalable to arbitrarily many input channels. We demonstrate its performance in learning frequency biases in the input as well as correlations between different input channels. Efficient computation of time-correlations requires a highly non-linear activation function. The resource requirements of a non-linear activation function are discussed. In addition to the thermodynamically consistent model of the CN, we also propose a biologically plausible version that could be engineered in a synthetic biology context.
△ Less
Submitted 28 September, 2020;
originally announced September 2020.
-
Minimal spiking neuron for solving multi-label classification tasks
Authors:
Jakub Fil,
Dominique Chu
Abstract:
The Multi-Spike Tempotron (MST) is a powerful single spiking neuron model that can solve complex supervised classification tasks. While powerful, it is also internally complex, computationally expensive to evaluate, and not suitable for neuromorphic hardware. Here we aim to understand whether it is possible to simplify the MST model, while retaining its ability to learn and to process information.…
▽ More
The Multi-Spike Tempotron (MST) is a powerful single spiking neuron model that can solve complex supervised classification tasks. While powerful, it is also internally complex, computationally expensive to evaluate, and not suitable for neuromorphic hardware. Here we aim to understand whether it is possible to simplify the MST model, while retaining its ability to learn and to process information. To this end, we introduce a family of Generalised Neuron Models (GNM) which are a special case of the Spike Response Model and much simpler and cheaper to simulate than the MST. We find that over a wide range of parameters the GNM can learn at least as well as the MST. We identify the temporal autocorrelation of the membrane potential as the single most important ingredient of the GNM which enables it to classify multiple spatio-temporal patterns. We also interpret the GNM as a chemical system, thus conceptually bridging computation by neural networks with molecular information processing. We conclude the paper by proposing alternative training approaches for the GNM including error trace learning and error backpropagation.
△ Less
Submitted 5 March, 2020;
originally announced March 2020.
-
Multi$^{\mathbf{3}}$Net: Segmenting Flooded Buildings via Fusion of Multiresolution, Multisensor, and Multitemporal Satellite Imagery
Authors:
Tim G. J. Rudner,
Marc Rußwurm,
Jakub Fil,
Ramona Pelich,
Benjamin Bischke,
Veronika Kopackova,
Piotr Bilinski
Abstract:
We propose a novel approach for rapid segmentation of flooded buildings by fusing multiresolution, multisensor, and multitemporal satellite imagery in a convolutional neural network. Our model significantly expedites the generation of satellite imagery-based flood maps, crucial for first responders and local authorities in the early stages of flood events. By incorporating multitemporal satellite…
▽ More
We propose a novel approach for rapid segmentation of flooded buildings by fusing multiresolution, multisensor, and multitemporal satellite imagery in a convolutional neural network. Our model significantly expedites the generation of satellite imagery-based flood maps, crucial for first responders and local authorities in the early stages of flood events. By incorporating multitemporal satellite imagery, our model allows for rapid and accurate post-disaster damage assessment and can be used by governments to better coordinate medium- and long-term financial assistance programs for affected areas. The network consists of multiple streams of encoder-decoder architectures that extract spatiotemporal information from medium-resolution images and spatial information from high-resolution images before fusing the resulting representations into a single medium-resolution segmentation map of flooded buildings. We compare our model to state-of-the-art methods for building footprint segmentation as well as to alternative fusion approaches for the segmentation of flooded buildings and find that our model performs best on both tasks. We also demonstrate that our model produces highly accurate segmentation maps of flooded buildings using only publicly available medium-resolution data instead of significantly more detailed but sparsely available very high-resolution data. We release the first open-source dataset of fully preprocessed and labeled multiresolution, multispectral, and multitemporal satellite images of disaster sites along with our source code.
△ Less
Submitted 4 December, 2018;
originally announced December 2018.
