Free

IsraelClouds AI-ML Forum #8

Event Information

Share this event

Date and Time

Location

Location

Natural Intelligence

6-8 Totseret ha-Arets Street

floor 23

Tel Aviv-Yafo, israel 6744129

Israel

View Map

Event description
Dimensionality Reduction | Voice Activity Detection

About this Event

Please join us to another fascinating evening about Dimensionality Reduction & Voice Activity Detection.

The event is kindly Hosted by Natural Intelligence.

Schedule:

18:30 - Gathering and mingling

18:45 - Welcoming

18:50 – Lecture #1: Dimensionality reduction: theoretical perspective on practical measures

19:35 – Lecture #2: Voice Activity Detection for Transient Noisy Environment Based on Diffusion Nets

20:20 - Sayonara

1st Session: Dimensionality reduction: theoretical perspective on practical measures

Metric dimensionality reduction, where data from a general metric is mapped into low dimensional space, is often used as a first step before applying machine learning algorithms. In almost all these applications the quality of the embedding is measured by various average-case criteria. Metric dimensionality reduction has also been studied in Math and TCS, within the field of metric embedding. Yet, the vast majority of theoretical research has been devoted to analyzing the worst-case behaviour of embeddings and therefore has little relevance to practical settings. I will talk about the theory and practice of metric dimensionality reduction, focusing on the rigorous analysis of the most common average-case measurement criteria used in practice. I will also present the results of empirical experiments that demonstrate that the theoretical results are exhibited in practical settings.

2nd Session: Voice Activity Detection for Transient Noisy Environment Based on Diffusion Nets

We address voice activity detection in acoustic environments of transients and stationary noises, which often occur in real life scenarios. We exploit unique spatial patterns of speech and non-speech audio frames by independently learning their underlying geometric structure. This process is done through a deep encoder-decoder based neural network architecture. This structure involves an encoder that maps spectral features with temporal information to their low- dimensional representations, which are generated by applying the diffusion maps method. The encoder feeds a decoder that maps the embedded data back into the high-dimensional space. A deep neural network, which is trained to separate speech from non-speech frames, is obtained by concatenating the decoder to the encoder, resembling the known Diffusion nets architecture. Experimental results show enhanced performance compared to competing voice activity detection methods. The improvement is achieved in both accuracy, robustness and generalization ability. Our model performs in a real-time manner and can be integrated into audio-based communication systems. We also present a batch algorithm which obtains an even higher accuracy for off-line applications.

Who can join us?

This session is for people with advanced background in the field of machine learning. Some of the content will be technical and academic-oriented.

About AI-BLOG:

AI-BLOG, the first and leading Hebrew blog on machine learning, deep learning, data science, and AI.

AI-Blog Community is about people that help each other to become more professional. It’s the place to learn and to share algorithms, methods and papers summaries between members.

www.ai-blog.co.il

About the lecturer - Nova Fandina:

Nova is a Phd student in the Computer Science Department at the Hebrew University. She works in the fields of approximation and metric embedding theory. Her main research agenda involves bridging the gap between theoretical investigations and practical applications. In particular, Nova believes that various phenomena observed in practice can and should be explained rigorously.

Nova is also highly engaged and committed to education, teaching and encouraging young students to fully appreciate the beauty of math

About the lecturer - Amir Ivry:

Amir Ivry received the B.Sc. degree in electrical engineering from the Technion, where he is currently working toward his PhD. degree (direct track) in electrical engineering. His research interests include deep learning for audiobased applications, such as voice activity detection, speech enhancement, and dereverberation. Amir is the recipient of the Jacobs Fellowship for Excellent Graduate Students for two years straight - 2019 and 2020. Amir is a full-time researcher and algorithms developer, and leads the artificial intelligence research group. Also, Amir is the author of 6 academic papers published in IEEE.

Share with friends

Date and Time

Location

Natural Intelligence

6-8 Totseret ha-Arets Street

floor 23

Tel Aviv-Yafo, israel 6744129

Israel

View Map

Save This Event

Event Saved