Distributed Learning of CNNs on Heterogeneous CPU/GPU Architectures

The convolutional neural networks (CNNs) have proven to be powerful classification tools in tasks that range from check reading to medical diagnosis, reaching close to human perception, and in some cases surpassing it. However, the problems to solve are becoming larger and more complex, which translates to larger CNNs, leading to longer training times that not even the adoption of Graphics Processing Units (GPUs) could keep up to. This problem is partially solved by using more processing units and distributed training methods that are offered by several frameworks dedicated to neural network training, such as Caffe, Torch, or TensorFlow. However, these techniques do not take full advantage of the possible parallelization offered by CNNs and the cooperative use of heterogeneous devices with different processing capabilities, clock speeds, memory size, among others. This paper presents a new method for the parallel training of CNNs where only the convolutional layer is distributed. The paper analyzes the influence of network size, bandwidth, batch size, number of devices, including their processing capabilities, and other parameters. Results show that this technique is capable of diminishing the training time without affecting the classification performance for both CPUs and GPUs. For the CIFAR-10 dataset, using a CNN with two convolutional layers, and 500 and 1500 kernels, respectively, best speedups achieve 3.28 x using four CPUs and 2.45 x with three GPUs. Larger datasets will certainly require more than 60-90% of processing time calculating convolutions, and speedups will tend to increase accordingly.

Bringing letters to life: handwriting with haptic-enabled tangible robots

Thibault Lucien Christian Asselborn, Wafa Monia Benkaouar Johal, Pierre Dillenbourg, Arzu Güneysu Özgür, Khalil Mrini, Ayberk Özgür, Elmira Yadollahi

In this paper, we present a robotic approach to improve the teaching of handwriting using the tangible, haptic-enabled and classroom-friendly Cellulo robots. Our efforts presented here are in line with the philosophy of the Cellulo platform: we aim to create a ready-to-use tool (i.e. a set of robot-assisted activities) to be used for teaching handwriting, one that is to coexist harmoniously with traditional tools and will contribute new added values to the learning process, complementing existing teaching practices. To maximize our potential contributions to this learning process, we focus on two promising aspects of handwriting: the visual perception and the visual-motor coordination. These two aspects enhance in particular two sides of the representation of letters in the mind of the learner: the shape of the letter (the grapheme) and the way it is drawn, namely the dynamics of the letter (the ductus). With these two aspects in mind, we do a detailed content analysis for the process of learning the representation of letters, which leads us to discriminate the specific skills involved in letter representation. We then compare our robotic method with traditional methods as well as with the combination of the two methods, in order to discover which of these skills can benefit from the use of Cellulo. As handwriting is taught from age 5, we conducted our experiments with 17 five-year-old children in a public school. Results show a clear potential of our robot-assisted learning activities, with a visible improvement in certain skills of handwriting, most notably in creating the ductus of the letters, discriminating a letter among others and in the average handwriting speed. Moreover, we show that the benefit of our learning activities to the handwriting process increases when it is used after traditional learning methods. These results lead to the initial insights into how such a tangible robotic learning technology may be used to create cost-effective collaborative scenarios for the learning of handwriting.

ACM2018

Brain signals of a Surprise-Actor-Critic model: Evidence for multiple learning modules in human decision making

Johanni Michael Brea, Wulfram Gerstner, Marco Philipp Lehmann, Vasiliki Liakoni, Alireza Modirshanechi, Kerstin Preuschoff

Learning how to reach a reward over long series of actions is a remarkable capability of humans, and potentially guided by multiple parallel learning modules. Current brain imaging of learning modules is limited by (i) simple experimental paradigms, (ii) entanglement of brain signals of different learning modules, and (iii) a limited number of computational models considered as candidates for explaining behavior. Here, we address these three limitations and (i) introduce a complex sequential decision making task with surprising events that allows us to (ii) dissociate correlates of reward prediction errors from those of surprise in functional magnetic resonance imaging (fMRI); and (iii) we test behavior against a large repertoire of model-free, model-based, and hybrid reinforcement learning algorithms, including a novel surprise-modulated actor-critic algorithm. Surprise, derived from an approximate Bayesian approach for learning the world-model, is extracted in our algorithm from a state prediction error. Surprise is then used to modulate the learning rate of a model-free actor, which itself learns via the reward prediction error from model-free value estimation by the critic. We find that action choices are well explained by pure model-free policy gradient, but reaction times and neural data are not. We identify signatures of both model-free and surprise-based learning signals in blood oxygen level dependent (BOLD) responses, supporting the existence of multiple parallel learning modules in the brain. Our results extend previous fMRI findings to a multi-step setting and emphasize the role of policy gradient and surprise signalling in human learning.

ACADEMIC PRESS INC ELSEVIER SCIENCE2022

Sparse and Low-rank Modeling for Automatic Speech Recognition

Pranay Dighe

This thesis deals with exploiting the low-dimensional multi-subspace structure of speech towards the goal of improving acoustic modeling for automatic speech recognition (ASR). Leveraging the parsimonious hierarchical nature of speech, we hypothesize that whenever a speech signal is measured in a high-dimensional feature space, the true class information is embedded in low-dimensional subspaces whereas noise is scattered as random high-dimensional erroneous estimations in the features. In this context, the contribution of this thesis is twofold: (i) identify sparse and low-rank modeling approaches as excellent tools for extracting the class-specific low-dimensional subspaces in speech features, and (ii) employ these tools under novel ASR frameworks to enrich the acoustic information present in the speech features towards the goal of improving ASR. Techniques developed in this thesis focus on deep neural network (DNN) based posterior features which, under the sparse and low-rank modeling approaches, unveil the underlying class-specific low-dimensional subspaces very elegantly. In this thesis, we tackle ASR tasks of varying difficulty, ranging from isolated word recognition (IWR) and connected digit recognition (CDR) to large-vocabulary continuous speech recognition (LVCSR). For IWR and CDR, we propose a novel \textit{Compressive Sensing} (CS) perspective towards ASR. Here exemplar-based speech recognition is posed as a problem of recovering sparse high-dimensional word representations from compressed low-dimensional phonetic representations. In the context of LVCSR, this thesis argues that albeit their power in representation learning, DNN based acoustic models still have room for improvement in exploiting the \textit{union of low-dimensional subspaces} structure of speech data. Therefore, this thesis proposes to enhance DNN posteriors by projecting them onto the manifolds of the underlying classes using principal component analysis (PCA) or compressive sensing based dictionaries. Projected posteriors are shown to be more accurate training targets for learning better acoustic models, resulting in improved ASR performance. The proposed approach is evaluated on both close-talk and far-field conditions, confirming the importance of sparse and low-rank modeling of speech in building a robust ASR framework. Finally, the conclusions of this thesis are further consolidated by an information theoretic analysis approach which explicitly quantifies the contribution of proposed techniques in improving ASR.