Université de Montréal

About: Université de Montréal is a education organization based out in Montreal, Quebec, Canada. It is known for research contribution in the topics: Population & Context (language use). The organization has 45641 authors who have published 100476 publications receiving 4004007 citations. The organization is also known as: University of Montreal & UdeM.

Mutual Information Neural Estimation.

Abstract: We argue that the estimation of mutual information between high dimensional continuous random variables can be achieved by gradient descent over neural networks. We present a Mutual Information Neural Estimator (MINE) that is linearly scalable in dimensionality as well as in sample size, trainable through back-prop, and strongly consistent. We present a handful of applications on which MINE can be used to minimize or maximize mutual information. We apply MINE to improve adversarially trained generative models. We also use MINE to implement the Information Bottleneck, applying it to supervised classification; our results demonstrate substantial improvement in flexibility and performance in these settings.

Distributed processing of pain and vibration by the human brain.

Abstract: Pain is a diverse sensory and emotional experience that likely involves activation of numerous regions of the brain. Yet, many of these areas are also implicated in the processing of nonpainful somatosensory information. In order to better characterize the processing of pain within the human brain, activation produced by noxious stimuli was compared with that produced by robust innocuous stimuli. Painful heat (47-48 degrees C), nonpainful vibratory (110 Hz), and neutral control (34 degrees C) stimuli were applied to the left forearm of right-handed male subjects. Activation of regions within the diencephalon and telencephalon was evaluated by measuring regional cerebral blood flow using positron emission tomography (15O-water-bolus method). Painful stimulation produced contralateral activation in primary and secondary somatosensory cortices (SI and SII), anterior cingulate cortex, anterior insula, the supplemental motor area of the frontal cortex, and thalamus. Vibrotactile stimulation produced activation in contralateral SI, and bilaterally in SII and posterior insular cortices. A direct comparison of pain and vibrotactile stimulation revealed that both stimuli produced activation in similar regions of SI and SII, regions long thought to be involved in basic somatosensory processing. In contrast, painful stimuli were significantly more effective in activating the anterior insula, a region heavily linked with both somatosensory and limbic systems. Such connections may provide one route through which nociceptive input may be integrated with memory in order to allow a full appreciation of the meaning and dangers of painful stimuli. These data reveal that pain-related activation, although predominantly contralateral in distribution, is more widely dispersed across both cortical and thalamic regions than that produced during innocuous vibrotactile stimulation. This distributed cerebral activation reflects the complex nature of pain, involving discriminative, affective, autonomic, and motoric components. Furthermore, the high degree of interconnectivity among activated regions may account for the difficulty of eliminating pathological pain with discrete CNS lesions.

Complexity of automaton identification from given data

Abstract: The question of whether there is an automaton with n states which agrees with a finite set D of data is shown to be NP-complete, although identification-in-the-limit of finite automata is possible in polynomial time as a function of the size of D. Necessary and sufficient conditions are given for D to be realizable by an automaton whose states are reachable from the initial state by a given set T of input strings. Although this question is also NP-complete, these conditions suggest heuristic approaches. Even if a solution to this problem were available, it is shown that finding a minimal set T does not necessarily give the smallest possible T.

FGF stimulation of the Erk1/2 signalling cascade triggers transition of pluripotent embryonic stem cells from self-renewal to lineage commitment

Abstract: Pluripotent embryonic stem (ES) cells must select between alternative fates of self-replication and lineage commitment during continuous proliferation. Here, we delineate the role of autocrine production of fibroblast growth factor 4 (Fgf4) and associated activation of the Erk1/2 (Mapk3/1) signalling cascade. Fgf4 is the major stimulus activating Erk in mouse ES cells. Interference with FGF or Erk activity using chemical inhibitors or genetic ablations does not impede propagation of undifferentiated ES cells. Instead, such manipulations restrict the ability of ES cells to commit to differentiation. ES cells lacking Fgf4 or treated with FGF receptor inhibitors resist neural and mesodermal induction, and are refractory to BMP-induced non-neural differentiation. Lineage commitment potential of Fgf4-null cells is restored by provision of FGF protein. Thus, FGF enables rather than antagonises the differentiation activity of BMP. The key downstream role of Erk signalling is revealed by examination of Erk2-null ES cells, which fail to undergo either neural or mesodermal differentiation in adherent culture, and retain expression of pluripotency markers Oct4, Nanog and Rex1. These findings establish that Fgf4 stimulation of Erk1/2 is an autoinductive stimulus for naive ES cells to exit the self-renewal programme. We propose that the Erk cascade directs transition to a state that is responsive to inductive cues for germ layer segregation. Consideration of Erk signalling as a primary trigger that potentiates lineage commitment provides a context for reconciling disparate views on the contribution of FGF and BMP pathways during germ layer specification in vertebrate embryos.