We study the problem of estimating ap-dimensional s-sparse vector in a linear model with Gaussian design and additivenoise.

Adder Neural Networks (ANNs) which only contain additions bring us a new way of developing deep neural networks with low energy consumption. Unfortunately, there is an accuracy drop when replacing all convolution filters by adder filters. The main reason here is the optimization difficulty of ANNs using $\ell_1$-norm, in which the estimation of gradient in back propagation is inaccurate. In this paper, we present a novel method for further improving the performance of ANNs without increasing the trainable parameters via a progressive kernel based knowledge distillation (PKKD) method. A convolutional neural network (CNN) with the same architecture is simultaneously initialized and trained as a teacher network, features and weights of ANN and CNN will be transformed to a new space to eliminate the accuracy drop. The similarity is conducted in a higher-dimensional space to disentangle the difference of their distributions using a kernel based method. Finally, the desired ANN is learned based on the information from both the ground-truth and teacher, progressively. The effectiveness of the proposed method for learning ANN with higher performance is then well-verified on several benchmarks. For instance, the ANN-50 trained using the proposed PKKD method obtains a 76.8\% top-1 accuracy on ImageNet dataset, which is 0.6\% higher than that of the ResNet-50.

Most recurrent neural networks (RNNs) do not include a fundamental constraint of real neural circuits: Dale's Law, which implies that neurons must be excitatory (E) or inhibitory (I). Dale's Law is generally absent from RNNs because simply partitioning a standard network's units into E and I populations impairs learning. However, here we extend a recent feedforward bio-inspired EI network architecture, named Dale's ANNs, to recurrent networks, and demonstrate that good performance is possible while respecting Dale's Law. This begs the question: What makes some forms of EI network learn poorly and others learn well? And, why does the simple approach of incorporating Dale's Law impair learning? Historically the answer was thought to be the sign constraints on EI network parameters, and this was a motivation behind Dale's ANNs. However, here we show the spectral properties of the recurrent weight matrix at initialisation are more impactful on network performance than sign constraints. We find that simple EI partitioning results in a singular value distribution that is multimodal and dispersed, whereas standard RNNs have an unimodal, more clustered singular value distribution, as do recurrent Dale's ANNs. We also show that the spectral properties and performance of partitioned EI networks are worse for small networks with fewer I units, and we present normalised SVD entropy as a measure of spectrum pathology that correlates with performance.

We consider the problem in Electrical Impedance Tomography (EIT) of identifying one or multiple inclusions in a background-conducting body $\Omega\subset\mathbb{R}^2$, from the knowledge of a finite number of electrostatic measurements taken on its boundary $\partial\Omega$ and modelled by the Dirichlet-to-Neumann (D-N) matrix. Once the presence of one inclusion in $\Omega$ is established, our model, combined with the machine learning techniques of Artificial Neural Networks (ANN) and Support Vector Machines (SVM), may be used to determine the size of the inclusion, the presence of multiple inclusions, and also that of anisotropy within the inclusion(s). Utilising both real and simulated datasets within a 16-electrode setup, we achieve a high rate of inclusion detection and show that two measurements are sufficient to achieve a good level of accuracy when predicting the size of an inclusion. This underscores the substantial potential of integrating machine learning approaches with the more classical analysis of EIT and the inverse inclusion problem to extract critical insights, such as the presence of anisotropy.

Anthony Bazhenov1,, - Pahan Dewasurendra1,, - Giri Krishnan2, and - Jean Erik Delanois2,3 1Del Norte High School, 16601 Nighthawk Ln, San Diego, CA 92127 2Department of Medicine, University of California, San Diego 3Department of Computer Science & Engineering, University of California, San Diego 9500 Gilman Drive, La Jolla, CA 92092 ABSTRACTDuring sleep, replay of recently learned memories along with relevant old memories enables the network to form The performance of artificial neural networks (ANNs) stable long-term memory representations [5] and reduces degrades when training data are limited or imbalanced. The idea of replay In contrast, the human brain can learn quickly from justhas been explored in machine learning to enable contina few examples. Here, we investigated the role of sleep inual learning. However, spontaneous unsupervised replay improving the performance of ANNs trained with limitedfound in the biological brain and implemented here is significantly different compared to explicit replay of pastdata on the MNIST and Fashion MNIST datasets. Sleep inputs implemented in machine learning rehearsal meth-was implemented as an unsupervised phase with local Hebbiantypelearningrules.

This research aims at comparative analysis of shear strength prediction at slab-column connection, unifying machine learning, design codes and Finite Element Analysis. Current design codes (CDCs) of ACI 318-19 (ACI), Eurocode 2 (EC2), Compressive Force Path (CFP) method, Feed Forward Neural Network (FNN) based Artificial Neural Network (ANN), PSO-based FNN (PSOFNN), and BAT algorithm-based BATFNN are used. The study is complemented with FEA of slab for validating the experimental results and machine learning predictions.In the case of hybrid models of PSOFNN and BATFNN, mean square error is used as an objective function to obtain the optimized values of the weights, that are used by Feed Forward Neural Network to perform predictions on the slab data. Seven different models of PSOFNN, BATFNN, and FNN are trained on this data and the results exhibited that PSOFNN is the best model overall. PSOFNN has the best results for SCS=1 with highest value of R as 99.37% and lowest of MSE, and MAE values of 0.0275%, and 1.214% respectively which are better than the best FNN model for SCS=4 having the values of R, MSE, and MAE as 97.464%, 0.0492%, and 1.43%, respectively.

The concept of Rao-Blackwellization is employed to improve predictions of artificial neural networks by physical information. The error norm and the proof of improvement are transferred from the original statistical concept to a deterministic one, using sufficient information on physics-based conditions. The proposed strategy is applied to material modeling and illustrated by examples of the identification of a yield function, elasto-plastic steel simulations, the identification of driving forces for quasi-brittle damage and rubber experiments. Sufficient physical information is employed, e.g., in the form of invariants, parameters of a minimization problem, dimensional analysis, isotropy and differentiability. It is proven how intuitive accretion of information can yield improvement if it is physically sufficient, but also how insufficient or superfluous information can cause impairment. Opportunities for the improvement of artificial neural networks are explored in terms of the training data set, the networks' structure and output filters. Even crude initial predictions are remarkably improved by reducing noise, overfitting and data requirements.

Despite the wild success of ChatGPT and other large language models, the artificial neural networks (ANNs) that underpin these systems might be on the wrong track. For one, ANNs are "super power-hungry," said Cornelia Fermüller, a computer scientist at the University of Maryland. "And the other issue is [their] lack of transparency." Such systems are so complicated that no one truly understands what they're doing, or why they work so well. This, in turn, makes it almost impossible to get them to reason by analogy, which is what humans do -- using symbols for objects, ideas and the relationships between them. Each neuron receives inputs, performs computations and produces outputs.

Spiking Neural Networks (SNNs) have received extensive academic attention due to the unique properties of low power consumption and high-speed computing on neuromorphic chips. Among various training methods of SNNs, ANN-SNN conversion has shown the equivalent level of performance as ANNs on large-scale datasets. However, unevenness error, which refers to the deviation caused by different temporal sequences of spike arrival on activation layers, has not been effectively resolved and seriously suffers the performance of SNNs under the condition of short time-steps. In this paper, we make a detailed analysis of unevenness error and divide it into four categories. We point out that the case of the ANN output being zero while the SNN output being larger than zero accounts for the largest percentage. Based on this, we theoretically prove the sufficient and necessary conditions of this case and propose an optimization strategy based on residual membrane potential to reduce unevenness error. The experimental results show that the proposed method achieves state-of-the-art performance on CIFAR-10, CIFAR-100, and ImageNet datasets. For example, we reach top-1 accuracy of 64.32\% on ImageNet with 10-steps. To the best of our knowledge, this is the first time ANN-SNN conversion can simultaneously achieve high accuracy and ultra-low-latency on the complex dataset. Code is available at https://github.com/hzc1208/ANN2SNN\_SRP.