1,720,982 research outputs found
A systematic assessment of deep learning models for molecule generation
Full text linkIn recent years the scientific community has devoted much effort in the development of deep learning models for the generation of new molecules with desirable properties (i.e. drugs). This has produced many proposals in literature. However, a systematic comparison among the different VAE methods is still missing. For this reason, we propose an extensive testbed for the evaluation of generative models for drug discovery, and we present the results obtained by many of the models proposed in literature
A better loss for visual-textual grounding
No full textGiven a textual phrase and an image, the visual grounding problem is the task of locating the content of the image referenced by the sentence. It is a challenging task that has several real-world applications in human-computer interaction, image-text reference resolution, and video-text reference resolution. In the last years, several works have addressed this problem by proposing more and more large and complex models that try to capture visual-textual dependencies better than before. These models are typically constituted by two main components that focus on how to learn useful multi-modal features for grounding and how to improve the predicted bounding box of the visual mention, respectively. Finding the right learning balance between these two sub-tasks is not easy, and the current models are not necessarily optimal with respect to this issue. In this work, we propose a loss function based on bounding boxes classes probabilities that: (i) improves the bounding boxes selection; (ii) improves the bounding boxes coordinates prediction. Our model, although using a simple multi-modal feature fusion component, is able to achieve a higher accuracy than state-of-the-art models on two widely adopted datasets, reaching a better learning balance between the two sub-tasks mentioned above
A better loss for visual-textual grounding
Full text linkGiven a textual phrase and an image, the visual grounding problem is the task of locating the content of the image referenced by the sentence. It is a challenging task that has several real-world applications in human-computer interaction, image-text reference resolution, and video-text reference resolution. In the last years, several works have addressed this problem by proposing more and more large and complex models that try to capture visual-textual dependencies better than before. These models are typically constituted by two main components that focus on how to learn useful multi-modal features for grounding and how to improve the predicted bounding box of the visual mention, respectively. Finding the right learning balance between these two sub-tasks is not easy, and the current models are not necessarily optimal with respect to this issue. In this work, we propose a loss function based on bounding boxes classes probabilities that: (i) improves the bounding boxes selection; (ii) improves the bounding boxes coordinates prediction. Our model, although using a simple multi-modal feature fusion component, is able to achieve a higher accuracy than state-of-the-art models on two widely adopted datasets, reaching a better learning balance between the two sub-tasks mentioned above
Inhibition and pleasure: economical risk-taking in the brain.
No full textWhen making decisions, people show different attitudes in risk-taking. Classically, individual differences have been investigated using personality tests but, recently, neuroscience methods are providing a novel point of view through which this aspect can be better understood. Here, we present a study in which participants play a gambling task by choosing between a first option that constantly yielded a small gain and a second option that provided either a larger gain or a loss. While participants performed the task, event-related potentials (ERPs) were recorded in order to investigate brain activity during the gambling task. Two groups were analysed post hoc: The first group was more risk-prone, while the second group was more risk-averse. The Feedback Related Negativity (FRN) occurring between 250 and 450 millisecond after decision outcomes, was differentially affected in the two groups. In addition, source analyses indicated that distinct brain areas are responsible for such a difference: Anterior Cingulate Cortex (ACC) is more activated in risk-prone group, while Dorso-Lateral Prefrontal Cortex (DLPFC) is more activated in risk-averse group. Our results show that risk-taking behaviour is related to differential activations in the brain. How the brain works may be used to predict the participants risk-attitude
Inducing disbelief in free will alters brain correlates of preconsciuous motor preparation: the brain minds whether we believe in free will or not
Full text link
The Detection and the Neural Correlates of Behavioral (Prior) Intentions
Full text linkPrior intentions are abstract mental representations that are believed to be the prime cause of our intentional actions. To date, only a few studies have focused on the possibility that single prior intentions could be identified in people's minds. Here, for the first time, we used the autobiographical Implicit Association Test (aIAT) in order to identify a specific prior intention on the basis of a pattern of associations derived from reaction times (Experiment 1). The aIAT is based on two critical blocks: the block associating intentions with true sentences (congruent block) gave rise to faster reaction times (RTs) than the block associating intentions with false sentences (incongruent block). Furthermore, when comparing intentions with hopes, it was revealed that the reported effect was intention-specific: The pattern of associations reflected a congruency effect when intentions and the logical category "True" were paired, but not when hopes and the "True" category were paired (Experiment 2). Finally, we investigated the neural bases of the congruency effect that leads to the identification of an intention (Experiment 3). We found a reduced late positive component (LPC) for the incongruent with respect to the congruent block, suggesting that the incongruent block needs additional resources of cognitive control with respect to the congruent block
- …
