Institute for Web Science and Technologies · Universität Koblenz - Landau
Institute WeST


Generating Counterfactual Images for Visual Question Answering by Editing Question-Critical Objects

17.06.21. While Visual Question Answering (VQA) systems improved significantly in recent years, they still tend to produce errors that are hard to reconstruct for human users. The lack of interpretability in black-box VQA models raises the necessity for discriminative explanations alongside the models’ outputs. This thesis aims at introducing a method to generate counterfactual images for an arbitrary VQA model. Given a question-image pair, the counterfactual generator should mask the question-critical objects in the image and then predict a minimal number of edits to the image such that the VQA model outputs a different answer. Thereby, the new image should contain semantically meaningful changes, be visually realistic, and remain unchanged in question-answer-irrelevant regions (e.g., the background). To the best of my knowledge, this is the first counterfactual image generator applied to VQA systems that does not apply edits to individual pixels but rather to a spatial mask without requiring additional manual annotations. [read more...]

Cryptocurrency Price Prediction Using Deep Learning and Social Media Sentiment Analysis

10.06.21. Despite the improvement that cryptocurrency price prediction had in recent years, these studies have not addressed the fact that cryptocurrency fundamentals are different from traditional markets which justifies the obtained results as being far from reliable. Unlike traditional markets, the hash rate of the mining process, the correlation between the cryptocurrency and global financial market, and public awareness are all factors that contribute to the price fluctuations of cryptocurrencies. This thesis aims at introducing a deep learning method for cryptocurrency price prediction on the next trading day using sentiment analysis from social media and search volumes in search engines. The data that will be used in this study is composed of the trading data of cryptocurrencies (time-series, bids, and asks), Twitter posts that will be analyzed, and assigned a weighting score depending on their sentiments and reach, along with search volumes for related terms to cryptocurrencies. Next, an LSTM model will use the collected data to predict the prices on the next trading day along with the confidence level in an effort to increase the overall accuracy. To the best of my knowledge, this is the first approach that uses deep learning with social media sentiments and reaches level along with search volumes to predict cryptocurrency prices. [read more...]

Cost Fairness for Blockchain-based Two Party Exchange Protocols

10.06.21. Existing blockchain-based fair exchange protocols usually neglect consideration of cost for assessing their fairness. However, in an environment with non-negligible cost, such as public blockchains, high or unexpected cost might be an obstacle for wide-spread adoption in business applications. We address this issue by defining cost fairness, which can be used to assess fair exchange protocols regarding its cost and discuss how it can be achieved in general and especially in context of blockchain. We show, that in an environment with non-negligible cost (such as a public blockchain), a fair exchange protocol cannot achieve full cost fairness without an external compensation mechanism. [read more...]

Preprocessing Argumentation Frameworks via Replacement Patterns

27.05.21. Diffrent abstract argumentation frameworks solvers are available today. While both specialized and constraint-based AF solvers have been developed, less attention has been so far put on the development of preprocessing and simplification techniques working directly on AFs. Dvorak et al. have developed different replacement patterns to preprocess and simplify AFs, which should lead to practical performance improvements and speedup for various AF solvers. This thesis aims to implement the presented replacment patterns with the help of the Tweety Project. Then it should be checked whether the results of Dvorak et al. are reproducible when using the AF solvers from ICCMA 2019. [read more...]

Towards Explainable Creativity: Tackling the Remote Association Test with Knowledge Graphs

20.05.21. Creative problem solving is one of the topics that interest both Cognitive Scientists and researchers in Artificial Intelligence. One aim of cognitive scientists is to build various A.I. systems that can solve creative problems and answer questions like 'How the human mind works while solving a creative task?' For Artificial Intelligence, creative problem-solving systems can help modeling agents solve complex problems with novel ideas. Various frameworks can perform computational creativity tests like the Remote Association Test (RAT) to measure AI systems' cognitive and problem-solving abilities. However, these frameworks cannot propose explanations for these solutions. In this talk, I propose a topic for my master's thesis: Building an AI system that can solve creativity tests and propose an explanation for these solutions. [read more...]

Older Entries