info.uvt.ro – Site FMI

Descriere1

>

Lorem ipsum dolor sit amet consectetur adipisicing elit. Maxime mollitia,
molestiae quas vel sint commodi repudiandae consequuntur voluptatum laborum
numquam blanditiis harum quisquam eius sed odit fugiat iusto fuga praesentium
optio, eaque rerum! Provident similique accusantium nemo autem. Veritatis
obcaecati tenetur iure eius earum ut molestias architecto voluptate aliquam
nihil, eveniet aliquid culpa officia aut! Impedit sit sunt quaerat, odit,
tenetur error, harum nesciunt ipsum debitis quas aliquid. Reprehenderit,
quia. Quo neque error repudiandae fuga? Ipsa laudantium molestias eos
sapiente officiis modi at sunt excepturi expedita sint? Sed quibusdam
recusandae alias error harum maxime adipisci amet laborum. Perspiciatis
minima nesciunt dolorem! Officiis iure rerum voluptates a cumque velit
quibusdam sed amet tempora. Sit laborum ab, eius fugit doloribus tenetur
fugiat, temporibus enim commodi iusto libero magni deleniti quod quam
consequuntur! Commodi minima excepturi repudiandae velit hic maxime
doloremque. Quaerat provident commodi consectetur veniam similique ad
earum omnis ipsum saepe, voluptas, hic voluptates pariatur est explicabo
fugiat, dolorum eligendi quam cupiditate excepturi mollitia maiores labore
suscipit quas? Nulla, placeat. Voluptatem quaerat non architecto ab laudantium
modi minima sunt esse temporibus sint culpa, recusandae aliquam numquam
totam ratione voluptas quod exercitationem fuga. Possimus quis earum veniam
quasi aliquam eligendi, placeat qui corporis!

This will close in 0 seconds

Descriere2

>

Lorem ipsum dolor sit amet consectetur adipisicing elit. Maxime mollitia,
molestiae quas vel sint commodi repudiandae consequuntur voluptatum laborum
numquam blanditiis harum quisquam eius sed odit fugiat iusto fuga praesentium
optio, eaque rerum! Provident similique accusantium nemo autem. Veritatis
obcaecati tenetur iure eius earum ut molestias architecto voluptate aliquam
nihil, eveniet aliquid culpa officia aut! Impedit sit sunt quaerat, odit,
tenetur error, harum nesciunt ipsum debitis quas aliquid. Reprehenderit,
quia. Quo neque error repudiandae fuga?

This will close in 0 seconds

Descriere3

>

Lorem ipsum dolor sit amet consectetur adipisicing elit. Maxime mollitia,
molestiae quas vel sint commodi repudiandae consequuntur voluptatum laborum
numquam blanditiis harum quisquam eius sed odit fugiat iusto fuga praesentium
optio, eaque rerum! Provident similique accusantium nemo autem. Veritatis
obcaecati tenetur iure eius earum ut molestias architecto voluptate aliquam
nihil, eveniet aliquid culpa officia aut! Impedit sit sunt quaerat, odit,
tenetur error, harum nesciunt ipsum debitis quas aliquid. Reprehenderit,
quia. Quo neque error repudiandae fuga?

This will close in 0 seconds

Cristina Soviany

AI based industry applications: A journey from cancer detection to market abuse detection

In this presentation, Cristina will walk us through her journey from leading the development of an innovative Machine Learning technology for cancer detection in ultrasound imaging to developing a breakthrough AI driven technology for the detection of market abuse in financial industry.

After finishing her PhD studies at Delft University of Technology in the Netherlands, Cristina worked for Philips Applied Technologies. In 2004, she moved to Belgium to join Advanced Medical Diagnostics (AMD), a start-up where she was leading a team of data scientists engineers, software developers and clinical research analysts to develop an innovative AI technology for prostate cancer detection in ultrasound imaging. Cristina will go through the machine learning technology principles and the methodologies used to develop the cancer detection technology and the Prostate Histoscanning TM medical device product. The critical need for absolute accuracy in cancer detection inspired her to transpose several years later similar AI and statistical based principles to detect fraud, the “cancer” of the financial industry.

After couple of years of initial development, Cristina has co-founded Features Analytics in 2014, and she is currently the CEO of this software and AI technology company based in Belgium. Features Analytics provides the next-generation Trade Surveillance Solutions for the detection of market abuse in trading in financial instruments. She will briefly introduce this topic to the audience, presenting the highlights of eyeDES®- the unique AI driven technology and platform developed by Features Analytics team under her leadership. Features Analytics is providing Trade Surveillance solutions to clients such as banks, exchanges and regulators in North America and Europe.

This will close in 0 seconds

Gabriel Istrate

Inteligenta Artificiala de încredere are nevoie de teoria jocurilor

În această prezentare vom încerca să argumentam faptul că noțiunile de teoria jocurilor pot juca un rol important în crearea de metode orientate către om în inteligența artificială de încredere. Drept exemplificare vom prezenta pe scurt rezultate (bazate pe o lucrare publicata in conferința TARK - Theoretical Aspects Rationality of Knowledge) asupra adecvării computaționale a conceptului de echilibru Kantian (Roemer 2019) și a unor concepte înrudite, precum cel de team reasoning (Bachrach 2006).

This will close in 0 seconds

Ștefan Trăușan-Matu

An Interdisciplinary Standpoint on Trustworthy Natural Language Processing

Trustworthy artificial intelligence applications that understand and generate natural language needs should follow a set of requirements among which two are ethicality and explainability. An interdisciplinary, dialogistic perspective for achieving these two aims will be presented, including fundamental research based on the polyphonic model and philosophy of language and artificial intelligence. Two implemented approaches developed at the Computers Department of the Politehnica University of Bucharest will introduced. A critical analysis of chatGPT and similar systems will be included.

This will close in 0 seconds

Decebal Constantin Mocanu

Dynamic Sparse Training: challenges and opportunities towards scalable, efficient, and sustainable AI

The talk starts with a quick overview of my research line. Next, it introduces one of the many challenges (the dense connectivity) which prevent us from having scalable artificial neural networks at both levels, cloud and edge computing. An emerging state-of-the-art possible solution is presented, i.e., sparse-to-sparse training with dynamic sparsity, which reduces the number of connections up to quadratically while persistently improving the performance. The discussion starts from the first sparse training works on complex Boltzmann machines and sparse evolutionary training in the typical single task (un)supervised learning. Further, it gradually introduces newer approaches in the more challenging contexts of continual or federated learning. Besides the fundamental theoretical novelty, some practical aspects, such as truly sparse implementations and deep learning energy efficiency are also considered and open for discussion.

This will close in 0 seconds

Andrei Ciortea

Hypermedia Multi-Agent Systems: Weaving a Web for People and Artificial Agents

The World Wide Web has emerged as the middleware of choice for most distributed systems. Recent standardization efforts for the Web of Things and Linked Data are now turning the Web into a homogeneous hypermedia fabric that interconnects everything — people, devices, information resources, abstract concepts, etc. The latest standards allow software agents not only to browse and query but also to observe and act on this hypermedia fabric. These recent developments unlock new practical use cases for engineering more sophisticated agents that can meet their design objectives through flexible autonomous use of this hypermedia fabric. Such autonomous agents have been studied to large extent in research on multi-agent systems (MAS). This talk explores the intersection of Web and MAS research to discuss a new class of Web-based MAS that are (i) aligned with the Web architecture to inherit the properties of the Web as a world-wide, open, and long-lived system, and (ii) transparent and accountable to support acceptance by people. The talk covers several demonstrators, including a prototypical system for industrial manufacturing developed at Siemens, and discusses some of the challenges and implications of designing a human-centered Web for artificial agents.

This will close in 0 seconds

Darian Onchiș

Explainability-driven model improvement for fault diagnosis trustworthiness

In this presentation, we overview two recent contributions to trustworthy applications of deep learning models in operational conditions. The first one is from 2021 and it is about the introduction of a compound stability-fit compensation index (SFC) for post-hoc surrogate explainability models. Since local explanations could become unstable, the SFC index is used as a quality indicator to accept an explanation. The index is checking the model from both a numerical perspective using the condition number and a statistical capacity through the R2 fit indicator. The second one is a novel cognitive neuro-symbolic deep learning architecture developed in 2022. The intrinsic explainability through the embeddings of symbolic reasoning into a convolutional neural network is used to create a hybrid deep learning model. One can query the satisfaction level of any defined formula or query the constraints while measuring the accuracy. Besides explainability, our model better avoids overfitting while also speeding up the training cycle.

This will close in 0 seconds

Bachelor Studies

Master Studies

News & Events

Details>>

May 05, 2022

Details>>

Alumni Community

Frequent questions

Partner

Contact FMCS

Contact FMCS

FMCS study programs

Fii la curent cu știrile UVT!

Newsletterul UVT

Linkuri utile

Press review

Descriere1

Descriere2

Descriere3

Cristina Soviany

Gabriel Istrate

Ștefan Trăușan-Matu

Decebal Constantin Mocanu

Andrei Ciortea

Darian Onchiș