SPIA Lectures

Detection of diabetic retinopathy based on a convolutional neural network using retinal fundus images

Diabetic retinopathy is one of the leading causes of blindness. Its damage is associated with the deterioration of blood vessels in retina. Progression of visual impairment may be cushioned or prevented if detected early, but diabetic retinopathy does not present symptoms prior to progressive loss of vision, and its late detection results in irreversible damages. Manual diagnosis is performed on retinal fundus images and requires experienced clinicians to detect and quantify the importance of several small details which makes this an exhaustive and time-consuming task. In this work, we attempt to develop a computer-assisted tool to classify medical images of the retina in order to diagnose diabetic retinopathy quickly and accurately. A neural network, with CNN architecture, identifies exudates, micro-aneurysms and hemorrhages in the retina image, by training with labeled samples provided by EyePACS, a free platform for retinopathy detection. The database consists of 35126 high-resolution retinal images taken under a variety of conditions. After training, the network shows a specificity of 93.65% and an accuracy of 83.68% on validation process.

January 5th, 09:15 - 10:00h

Manasses Antoni Mauricio Condori

MSc Student. CS UCSP

PDF Presentation

PRESENTACION_UCSP_02.pdf

Teaching Robots With Interactive Reinforcement Learning

We investigate learning approaches, more specifically interactive reinforcement learning to perform a domestic task. We use parent-like advice to explore two set-ups: agent-agent and human-agent interaction.

In this presentation, we show our contribution to knowledge in terms of studying the interplay of multi-modal interactive feedback and contextual affordances. Overall, we investigate which parameters influence the interactive reinforcement learning process and show that the apprenticeship of reinforcement learning agents can be sped up by means of interactive parent-like advice, multi-modal feedback, and affordances-driven environmental models.

January 5th, 10:00 - 10:45h

Francisco Cruz

https://www.inf.uni-hamburg.de/en/inst/ab/wtm/people/cruz.html

Extractive document summarization using complex network measurements

Due to the large amount of textual information available on Internet, there is an increasingly need to use techniques that find relevant and concise content in large volumes of information. A typical task devoted to the identification of relevant and informative sentences in documents is the so called extractive document summarization. In this presentation, we show our research in complex network concepts to devise an extractive document summarization method, which extracts the most relevant sentences from several textual sources. More specifically, texts are modeled as networks, where nodes represent sentences and the edges are set based on different criteria. Finally, the identification of relevant sentences is guided by the characterization of nodes via complex network measurements.

Jorge Andoni Valverde Tohalino

Universidade de São Paulo (USP)

January 5th, 10:45 - 11:30h

Learning to Play DOOM

In contrast to supervised learning, there are different problem settings in which we rarely observe labels and most of the information is expressed in terms of rewards and goals. Different instantiations of these type of problems show the need of learning by reinforcing rewards in the environment, just like a video game. In this talk, we will teach an agent to play DOOM with the goal of surviving in an environment with opponents and limited munitions supply.

Omar U. Florez (Research Scientist - Intel Labs California)

January 5th, 11:30 - 12:15h

Google Sites

Report abuse