Mehdi Mirza's research while affiliated with Université de Montréal and other places

Publications (13)

Article
Humans learn a predictive model of the world and use this model to reason about future events and the consequences of actions. In contrast to most machine predictors, we exhibit an impressive ability to generalize to unseen scenarios and reason intelligently in these settings. One important aspect of this ability is physical intuition(Lake et al.,...
Article
Full-text available
Theano is a Python library that allows to define, optimize, and evaluate mathematical expressions involving multi-dimensional arrays efficiently. Since its introduction, it has been one of the most used CPU and GPU mathematical compilers - especially in the machine learning community - and has shown steady performance improvements. Theano is being...
Conference Paper
The task of the emotion recognition in the wild (EmotiW) Challenge is to assign one of seven emotions to short video clips extracted from Hollywood style movies. The videos depict acted-out emotions under realistic conditions with a large degree of variation in attributes such as pose and illumination, making it worthwhile to explore approaches whi...
Article
Full-text available
We propose a new framework for estimating generative models via an adversarial process, in which we simultaneously train two models: a generative model G that captures the data distribution, and a discriminative model D that estimates the probability that a sample came from the training data rather than G. The training procedure for G is to maximiz...
Article
We propose a new framework for estimating generative models via an adversarial process, in which we simultaneously train two models: a generative model G that captures the data distribution, and a discriminative model D that estimates the probability that a sample came from the training data rather than G. The training procedure for G is to maximiz...
Article
Full-text available
Catastrophic forgetting is a problem faced by many machine learning models and algorithms. When trained on one task, then trained on a second task, many machine learning models "forget'' how to perform the first task. This is widely believed to be a serious problem for neural networks. Here, we investigate the extent to which the catastrophic forge...
Conference Paper
In this paper we present the techniques used for the University of Montréal's team submissions to the 2013 Emotion Recognition in the Wild Challenge. The challenge is to classify the emotions expressed by the primary human subject in short video clips extracted from feature length movies. This involves the analysis of video clips of acted scenes la...
Article
Full-text available
Pylearn2 is a machine learning research library. This does not just mean that it is a collection of machine learning algorithms that share a common API; it means that it has been designed for flexibility and extensibility in order to facilitate research projects that involve new or unusual use cases. In this paper we give a brief history of the lib...
Article
Full-text available
The ICML 2013 Workshop on Challenges in Representation Learning. 11http://deeplearning.net/icml2013-workshop-competition. focused on three challenges: the black box learning challenge, the facial expression recognition challenge, and the multimodal learning challenge. We describe the datasets created for these challenges and summarize the results o...
Article
Full-text available
We consider the problem of designing models to leverage a recently introduced approximate model averaging technique called dropout. We define a simple new model called maxout (so named because its output is the max of a set of inputs, and because it is a natural companion to dropout) designed to both facilitate optimization by dropout and improve t...
Conference Paper
We consider the problem of designing models to leverage a recently introduced approximate model averaging technique called dropout. We define a simple new model called maxout (so named because its output is the max of a set of inputs, and because it is a natural companion to dropout) designed to both facilitate optimization by dropout and improve t...
Article
We introduce the multi-prediction deep Boltzmann machine (MP-DBM). The MPDBM can be seen as a single probabilistic model trained to maximize a variational approximation to the generalized pseudolikelihood, or as a family of recurrent nets that share parameters and approximately solve different inference problems. Prior methods of training DBMs eith...
Conference Paper
We propose a semi-supervised approach to solve the task of emotion recognition in 2D face images using recent ideas in deep learning for handling the factors of variation present in data. An emotion classification algorithm should be both robust to (1) remaining variations due to the pose of the face in the image after centering and alignment, (2)...

Citations

... Therefore, Biswas et al. [41] used the smooth function to approximate the |x| function. They found a general approximation formula of the maximum function from the smooth approximation of the |x| function, which can smoothly approximate the general maxout [42] family, ReLU, leaky ReLU, or its variants, such as Swish, etc. In addition, the author also proves that the GELU function is a special case of the SMU. ...
... Another potential method to address data sparsity concerns is data augmentation using techniques such as Generative Adversarial Networks (GANs) (Goodfellow et al. 2014;Han et al. 2018;Salimans et al. 2016) to generate new samples (Donahue et al. 2018;Saito et al. 2018). GANs consists of two neural networks: a generative model (generator) and a discriminative model (discriminator) which are set to compete against each other in a zero-sum game. ...
... Therefore, the research focused strongly on face recognition, an active research area in recent years. In this category are the works proposed by Kahou et al. 2015, Kollias et al. 2015and Wei et al. 2017. Unfortunately, these solutions lose generality as they are strongly focused on the primary detection of the face without considering other aspects that make up the image. ...
... They use CNN based image classifiers taking as input an image of a block tower and returning a probability for the tower to fall. Lerer et al. (2016); Mirza et al. (2017) also include a decoding module to predict final positions of these blocks. Groth et al. (2018) investigate the ability of such a model to actively position shapes in stable tower configurations. ...
... The continuous rise of DL has propelled the growth of various open-source libraries, like Tensorflow [19], Keras [20], PyTorch [21], Theano [22], etc., for efficient data flow while implementing DL models. In this paper, we primarily focus on PyTorch and identified an implementation vulnerability typically responsible for Class-Leakage through timing side-channel 2 . ...
... Methods based on implicit optimization layers come under the category of Optimization-based Modeling architectures [31] and are well-studied for generic classification, and structured prediction tasks [32,33,34,35]. The most common way of training such models is through Unrolled Differentiation [36,37,38,33]. ...
... The algorithm, generative adversarial network (GAN), was first developed to improve the performance of deep learning networks in approximating intractable probabilistic computations [5]. The GAN algorithm consists of two competing networks, the generator network and the discriminator network. ...
... Wang et al. constructed an architecture similar to U-Net as an attention branch to highlight subtle local facial expression information [35]. The local representation was obtained by a multi-scale contractive convolutional network (CCNET) in [36]. A multi-layer network after CNN architecture was exploited in [37] to learn higher-level features for FER. ...
... Still, even several studies have found that concatenating multiple input variables increase the model performances (Kanou et al. 2013;Schouten et al. 2016), the general trend in our unimodal procedure suggests that there is no gained information from incorporating additional information. Thence, the a priori expected improvement in classification performance in the multimodal case over the best single modality measure did not concretise. ...
... Because our methods learn each task well and outperform the others with significant magnitudes. Moreover, some studies (De Lange et al., 2021;Goodfellow et al., 2013) practically investigated dropout in continual learning and they showed that dropout can reduce the catastrophic forgetting phenomenon. Meanwhile, according to our theoretical analyses, SVB is the most stable, therefore, the average LLPs of SVB reduce the least on both of the two datasets. ...