In modern deep learning, algorithmic choices (such as width, depth, and learning rate) are known to modulate nuanced resource tradeoffs. This work investigates how these complexities necessarily arise for feature learning in the presence of computational-statistical gaps. We begin by considering offline sparse parity learning, a supervised classification problem which admits a statistical query lower bound for gradient-based training of a multilayer perceptron. This lower bound can be interpreted as a: successful learning can only occur if one is sufficiently rich (large model), knowledgeable (large dataset), patient (many training iterations), or lucky (many random guesses). We show, theoretically and experimentally, that sparse initialization and increasing network width yield significant improvements in sample efficiency in this setting. Here, width plays the role of parallel search: it amplifies the probability of finding lottery ticket neurons, which learn sparse features more sample-efficiently. Finally, we show that the synthetic sparse parity task can be useful as a proxy for real problems requiring axis-aligned feature learning. We demonstrate improved sample efficiency on tabular classification benchmarks by using wide, sparsely-initialized MLP models; these networks sometimes outperform tuned random forests.

In this paper, we formalize the idea behind capsule nets of using a capsule vector rather than a neuron activation to predict the label of samples. To this end, we propose to learn a group of capsule subspaces onto which an input feature vector is projected. Then the lengths of resultant capsules are used to score the probability of belonging to different classes. We train such a Capsule Projection Network (CapProNet) by learning an orthogonal projection matrix for each capsule subspace, and show that each capsule subspace is updated until it contains input feature vectors corresponding to the associated class. With low dimensionality of capsule subspace as well as an iterative method to estimate the matrix inverse, only a small negligible computing overhead is incurred to train the network. Experiment results on image datasets show the presented network can greatly improve the performance of state-of-the-art Resnet backbones by $10-20\%$ with almost the same computing cost.

In modern deep learning, algorithmic choices (such as width, depth, and learning rate) are known to modulate nuanced resource tradeoffs. This work investigates how these complexities necessarily arise for feature learning in the presence of computational-statistical gaps. We begin by considering offline sparse parity learning, a supervised classification problem which admits a statistical query lower bound for gradient-based training of a multilayer perceptron. This lower bound can be interpreted as a multi-resource tradeoff frontier: successful learning can only occur if one is sufficiently rich (large model), knowledgeable (large dataset), patient (many training iterations), or lucky (many random guesses). We show, theoretically and experimentally, that sparse initialization and increasing network width yield significant improvements in sample efficiency in this setting.

This paper introduces an alternative to CNN based architectures being inspired by the recently proposed capsule networks. The authors proposed to replace the last layer of ResNet variants by a capsule projection network, thereby getting promising results on the CIFAR and SVHN datasets. However, the motivation for using a capsule projection layer is unclear even though the technique is straightforward and easy to implement with minor computational overhead. The main idea of the capsule projection layer is to project the input feature vector to some learnt capsule subspaces (one for each class in classification setting), which are then used to distinguish between the different classes in classification. The authors also show that this projection technique leads to computation of gradients which are orthogonal to the learnt subspace, enabling discovery of novel characteristics leading to improvement of the learnt subspace.

In recent years, the traditional feature engineering process for training machine learning models is being automated by the feature extraction layers integrated in deep learning architectures. In wireless networks, many studies were conducted in automatic learning of feature representations for domain-related challenges. However, most of the existing works assume some supervision along the learning process by using labels to optimize the model. In this paper, we investigate an approach to learning feature representations for wireless transmission clustering in a completely unsupervised manner, i.e. requiring no labels in the process. We propose a model based on convolutional neural networks that automatically learns a reduced dimensionality representation of the input data with 99.3% less components compared to a baseline principal component analysis (PCA). We show that the automatic representation learning is able to extract fine-grained clusters containing the shapes of the wireless transmission bursts, while the baseline enables only general separability of the data based on the background noise.

The purpose of this paper is to compare different learnable frontends in medical acoustics tasks. A framework has been implemented to classify human respiratory sounds and heartbeats in two categories, i.e. healthy or affected by pathologies. After obtaining two suitable datasets, we proceeded to classify the sounds using two learnable state-of-art frontends -- LEAF and nnAudio -- plus a non-learnable baseline frontend, i.e. Mel-filterbanks. The computed features are then fed into two different CNN models, namely VGG16 and EfficientNet. The frontends are carefully benchmarked in terms of the number of parameters, computational resources, and effectiveness. This work demonstrates how the integration of learnable frontends in neural audio classification systems may improve performance, especially in the field of medical acoustics. However, the usage of such frameworks makes the needed amount of data even larger. Consequently, they are useful if the amount of data available for training is adequately large to assist the feature learning process.

In this paper, we formalize the idea behind capsule nets of using a capsule vector rather than a neuron activation to predict the label of samples. To this end, we propose to learn a group of capsule subspaces onto which an input feature vector is projected. Then the lengths of resultant capsules are used to score the probability of belonging to different classes. We train such a Capsule Projection Network (CapProNet) by learning an orthogonal projection matrix for each capsule subspace, and show that each capsule subspace is updated until it contains input feature vectors corresponding to the associated class. With low dimensionality of capsule subspace as well as an iterative method to estimate the matrix inverse, only a small negligible computing overhead is incurred to train the network.

Recent years have seen a great interest in using deep architectures for feature learning from data. One drawback of the commonly used unsupervised deep feature learning methods is that for supervised or semi-supervised learning tasks, the information in the target variables are not used until the final stage when the classifier or regressor is trained on the learned features. This could lead to over-generalized features that are not competitive on the specific supervised or semi-supervised learning tasks. In this work, we describe a new learning method that combines deep feature learning on mixed labeled and unlabeled data sets. Specifically, we describe a weakly supervised learning method of a prior supervised convolutional stacked auto-encoders (PCSA), of which information in the target variables is represented probabilistically using a Gaussian Bernoulli restricted Boltzmann machine (RBM). We apply this method to the decoding problem of an ECoG based Brain Computer Interface (BCI) system. Our experimental results show that PCSA achieves significant improvement in decoding performance on benchmark data sets compared to the unsupervised feature learning as well as to the current state-of-the-art algorithms that are based on manually crafted features.