Bow-tie architectures in biological and artificial neural networks: Implications for network evolution and assay design

Hilliard, Seth; Mosoyan, Karen; Branciamore, Sergio; Gogoshin, Grigoriy; Zhang, Alvin; Simons, Diana L.; Rockne, Russell; Lee, Peter P.; Rodin, Andréi S.

doi:10.1016/j.isci.2023.106041

Cited by 4 publications

(4 citation statements)

References 51 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“… 2 ). Similarly, information processing has been reported to be robust against the increase in input/output nodes 38 . These findings suggest that the addition of inputs and outputs (i.e., goal matrix expansion) does not disturb the bow-tie structure, and can even promote its emergence (Fig.…”

Section: Discussionmentioning

confidence: 96%

See 1 more Smart Citation

Revisiting the evolution of bow-tie architecture in signaling networks

Itoh,

Kondo,

Aoki

et al. 2024

npj Syst Biol Appl

View full text Add to dashboard Cite

Bow-tie architecture is a layered network structure that has a narrow middle layer with multiple inputs and outputs. Such structures are widely seen in the molecular networks in cells, suggesting that a universal evolutionary mechanism underlies the emergence of bow-tie architecture. The previous theoretical studies have implemented evolutionary simulations of the feedforward network to satisfy a given input-output goal and proposed that the bow-tie architecture emerges when the ideal input-output relation is given as a rank-deficient matrix with mutations in network link intensities in a multiplicative manner. Here, we report that the bow-tie network inevitably appears when the link intensities representing molecular interactions are small at the initial condition of the evolutionary simulation, regardless of the rank of the goal matrix. Our dynamical system analysis clarifies the mechanisms underlying the emergence of the bow-tie structure. Further, we demonstrate that the increase in the input-output matrix reduces the width of the middle layer, resulting in the emergence of bow-tie architecture, even when evolution starts from large link intensities. Our data suggest that bow-tie architecture emerges as a side effect of evolution rather than as a result of evolutionary adaptation.

show abstract

Section: Discussionmentioning

confidence: 96%

“…Thus, a comparison of neural networks and the biological bow-tie architecture might yield interesting insights. For example, from this aspect, the information-theoretic principle 38 and classification ability 17 have been evaluated in signaling pathways.…”

Section: Discussionmentioning

confidence: 99%

Revisiting the evolution of bow-tie architecture in signaling networks

Itoh,

Kondo,

Aoki

et al. 2024

npj Syst Biol Appl

View full text Add to dashboard Cite

show abstract

“…The concept behind these networks draws inspiration from the functioning of the biological nervous system (BNS) in processing data and information to facilitate learning and knowledge creation. A pivotal aspect of this concept involves the development of novel structures for information processing systems [15]. This system comprises numerous interconnected processing units known as neurons, which collaborate to address problems and transmit information via electromagnetic communication.…”

Section: Introductionmentioning

confidence: 99%

Image Captioning by Using Bidirectional LSTM Neural Network

Hoseini,

Notash

2024

Preprint

View full text Add to dashboard Cite

Putting captions on images is an important phase in image processing and machine vision which divides an image into different regions. In the related captioning operation, similar attributes are extracted from each region and a single caption is opted for the title of the input for the relevant image. In the developed world, due to the considerable volume of images that have been available on the Internet, it is complicated to choose a suitable method for captioning images automatically and determining concepts correctly. In this study, a method based on a Bidirectional LSTM neural network with 14 layers is used for automatic captioning. The dataset is trained on ImageNet by MobileNet architecture. Due to the high volume of the data set consequently with the huge calculations that are required for captioning, MobileNet architecture has been applied to decrease the computational load and to boost the performance of the proposed method. in this study, the data set is flickr8k with a volume of data of more than 1.04 GB which was selected as one of the challenging data sets for captioning images through the Kaggle site. The proposed method is implemented with Python language by artificial intelligence package TensorFlow and Cross on Google Colab. The proposed Bidirectional LSTM neural network has been compared with five other models based on the LSTM neural network. So, the results of these comparisons have been evaluated for Precision, Accuracy, Recall, F score, and Loss Function. According to the hardware's limitations and the huge volume of calculations, the proposed model was able to show convincing performance compared to other models. The obtained results with an accuracy of 75.90% indicate the improvement of the accuracy as well as the reduction of the cost in the captioning process for the proposed model.

show abstract

“…Bow-tie architecture (BTA) is ubiquitous in biological systems (1)(2)(3). BTA consists of substantially smaller or simpler intermediate systems linking between much larger and more complex systems.…”

Section: Introductionmentioning

confidence: 99%

Non-Negative Connectivity Causes Bow-Tie Architecture in Neural Circuits

Liu,

Du,

Wong-Lin

et al. 2024

Preprint

View full text Add to dashboard Cite

Bow-tie or hourglass architecture is commonly found in biological neural networks. Recently, artificial neural networks with bow-tie architecture have been widely used in various machine-learning applications. However, it is unclear how bow-tie architecture in neural circuits can be formed. We address this by training multi-layer neural network models to perform classification tasks. We demonstrate that during network learning and structural changes, non-negative connections amplify error signals and quench neural activity particularly in the hidden layer, resulting in the emergence of the network's bow-tie architecture. We further show that such architecture has low wiring cost, robust to network size, and generalizable to different discrimination tasks. Overall, our work suggests a possible mechanism for the emergence of bow-tie neural architecture and its functional advantages.

show abstract

Bow-tie architectures in biological and artificial neural networks: Implications for network evolution and assay design

Cited by 4 publications

References 51 publications

Revisiting the evolution of bow-tie architecture in signaling networks

Revisiting the evolution of bow-tie architecture in signaling networks

Image Captioning by Using Bidirectional LSTM Neural Network

Non-Negative Connectivity Causes Bow-Tie Architecture in Neural Circuits

Contact Info

Product

Resources

About