Anđela Todorović

Information & Communications Technology

Machine Learning and AI Software Engineering

Niš, Central Serbia, Serbia

Passionate about inspiring others to create and achieve more in the field of intelligent computing. 🍀
Associate Software Engineer at Cubic Corporation.
2 x Bachelor student of Computer Science / Mathematics.
2 x Founder of an award-winning software solution .

Current sessions

Gene genealogies reconstruction on large-scale BALSAC data

The objective of this talk arises from the reconstruction of gene genealogies problem - from a given genotypes of sequence data from contemporary individuals and an extended pedigree of genealogical relationships among them, we have to decide and be very clear about what we should consider to be nodes information in genealogy tree according to kinds of dependencies among these nodes which should be considered to be edges of the tree.

This talk will present novel algorithms for advanced reconstruction of
the gene genealogy by performing inference on graphical models, specifically, implementing greedy and belief propagation algorithms on the succinct tree sequences.The implemented algorithms will in turn be integrated into the existing tskit and msprime libraries. The succinct tree sequence is a recently introduced encoding for recombinant ancestry that takes advantage of the correlations between adjacent trees. Considering that tree sequence data structure has the potential to hugely reduce storage and processing costs, it is a natural (if not the only) contender to tackle this problem.

Deep learning for sensor-based activity recognition

Human activity detection has seen immense growth in the last decade playing an important role in the field of pervasive computing. This popularity can be credited to its real-life applications primarily dealing with human-centric problems in general medicine. Many research attempts with data mining and machine learning techniques have been undergoing to accurately detect human activities for healthcare systems. This speech reviews some of the predictive machine learning algorithms and compares the accuracy and performances of these models.
Using CNNs we have managed to broadly categorize the target activities as moving (walking, walking upstairs downstairs) and sedentary (sitting, standing and lying). The output of the model in the form of contingency tables demonstrated a strong ability to separate between the activity categories and distinguished the separate activities in the categories with lower, but still notable success rates.

Advanced Deep Learning techniques in Semantic Analysis on large scale datasets

This speech reviews some of the latest approaches in the usage of advanced machine learning techniques on the large scale datasets, mainly for semantic analysis. The existing deep learning techniques are broadly classified for big data, and the architecture of models is principally based on deep belief networks and convolution neural networks. However, deep learning techniques have various limitations in processing big data with current techniques. One of the major reasons is processing the big and large amount of data in vector space. However, several sophisticated and optimized algorithms exist and are uttered to process data with high speed and classify feature learning with high accuracy, and they are to be presented.

Use of LSTMs for forecasting Server Requests

Machine learning, especially time-series data analysis, can help with the automatization of many tasks, especially in DevOps and SRE.
It has shown high-grade results with keeping the past data for an established reaction to something new.
It can also fix the issue and alert us on what is going to happen instead of having someone manually analyzing the prior results and attempting to preview the future.
LSTMs are a special type of recurrent neural networks (RNNs) that have shown a desired behavior on the time-series data.
The quick demo will show the results on predicting tendency over the next 24 hours by using time series data of the past 100 days of nginx request.

The usage of Artificial Intelligence in leveraging SMEs

Machine learning and deep learning use-cases are not restricted to large enterprises now. They are an exceptional way to achieve tremendous growth for SMEs as well, by allowing small businesses to effectively transfer conventional tasks to an innovative, yet fast approach.
Data generated in SMEs, especially time-series data, is very rich in information that can provide valuable insights on the current and future health of the production systems and the products they create. 
However, it is strongly recommended for small businesses to strategize properly when starting integrating AI if they want to reach potential consumers in the future.
By choosing an inadequate approach, much of the gathered data goes underutilized as traditional methods have limitations to effectively leverage multivariate trends and uncover new insights for improving operations.
This speech will consult good strategies for developing proper business models, as far as discuss the process of integrating machine learning as a tool in an existing product.

Improving advanced object detection algorithms using Content Based Image Retrieval

Content-based image retrieval utilizes representations of features that are automatically extracted from the
images, mainly perceptual properties, such as shape, color, texture, and spatial relationships to find similar-looking images.
The usage of CBIR techniques aids in resolving the problem of missing annotations from the extracted video frames that have more eminent noise and overall lower quality, by permitting them to inherit (to some extent and with lower confidence) the tags retrieved from the similar-looking frames, which will result in an overall increase of the trustworthiness of the object detection algorithms themselves.
This speech will review the implementation of a video content categorization model using the
YOLOv3 implementation pre-trained on the MS-COCO and fine-tuned on the custom dataset, followed by the presentation on the improvements in overall results by using the custom-built convolutional neural network for CBIR and tag suggestion.

Past and future events

Tech Conference Europe Spring Edition

16 Mar 2021
Prague, Hlavní město Praha, Czechia

Artificial Intelligence and I

11 Nov 2020
Belgrade, Central Serbia, Serbia

Annual Summit on Artificial Intelligence and Machine Learning

26 Feb - 27 Feb 2020
Dubai, United Arab Emirates

IT Space Conference

10 Dec 2019
Belgrade, Central Serbia, Serbia

DataFest Tbilisi 2019

13 Nov - 16 Nov 2019
Tbilisi, T'bilisi, Georgia

PyCon Balkan 2019

3 Oct - 5 Oct 2019
Belgrade, Central Serbia, Serbia

Google Inside Look

4 Sep - 6 Sep 2019
Munich, Bavaria, Germany

PICANTE Tech Conference Europe

3 Sep - 4 Sep 2019
Prague, Hlavní město Praha, Czechia

Digital Influencers in Action Forum

17 Jun - 19 Jun 2019
Budva, Montenegro

WeAreDevelopers World Congress

7 Jun - 8 Jun 2019
Berlin, Germany

Machine Learning Conference Serbia

24 May - 25 May 2019
Subotica, Vojvodina, Serbia

CodeCamp Timisoara

18 May 2019
Timişoara, Timiş, Romania

Machine Learning Meetup

14 May 2019
Niš, Central Serbia, Serbia

WeBiz Conference 7.0

29 Mar - 31 Mar 2019
Zrenjanin, Vojvodina, Serbia

Mobile World Congress 2019

25 Feb - 28 Feb 2019
Barcelona, Catalonia, Spain

Advanced Technology Days 14

4 Dec - 5 Dec 2018
Zagreb, City of Zagreb, Croatia

Upbeat forum on Blockchain and Social Innovation

4 Dec - 5 Dec 2018
Podgorica, Montenegro

Macedonian CodeCamp 2018

16 Nov 2018
Skopje, Grad Skopje, North Macedonia

PyCon Balkan 2018

16 Nov 2018
Belgrade, Central Serbia, Serbia

satRday Belgrade 2018

26 Oct 2018
Belgrade, Central Serbia, Serbia

School of AI

13 Oct 2018
Niš, Central Serbia, Serbia

DevFest Milano 2018

6 Oct 2018
Milan, Lombardy, Italy

Point conference 7.0

17 May - 19 May 2018
Sarajevo, Federation of B&H, Bosnia and Herzegovina

Mobile World Congress 2016

24 Feb - 28 Feb 2016
Barcelona, Catalonia, Spain