- 106
- 62 898
OLIVES at GATECH
United States
Приєднався 19 січ 2010
OLIVES at Georgia Tech
Omni Lab for Intelligent Visual Engineering and Science
ghassanalregib.info
Omni Lab for Intelligent Visual Engineering and Science
ghassanalregib.info
ICME'24 Tutorial on Robust Image Understanding: Explainability, Uncertainty, and Intervenability
Neural network driven applications like ChatGPT suffer from hallucinations where they confidently provide inaccurate information. A fundamental reason for this inaccuracy is the lack of robust measures that are applied on the underlying neural network predictions. In this tutorial, we identify and expound on three human-centric robustness measures, namely explainability, uncertainty, and intervenability, that every decision made by a neural network must be equipped and evaluated with. Explainability and uncertainty research fields are accompanied by a large body of literature that analyze decisions. Intervenability, on the other hand, has gained recent prominence due its inclusion in the GDPR regulations and a surge in prompting-based neural network architectures. In this tutorial, we connect all three fields using inference-based reliability assessment techniques to motivate robust image interpretation.
For the slides, please visit the following website:
alregib.ece.gatech.edu/icme-2024-tutorial/
For the slides, please visit the following website:
alregib.ece.gatech.edu/icme-2024-tutorial/
Переглядів: 78
Відео
WACV’24 Tutorial on Robustness at Inference: Explainability, Uncertainty, and Intervenability
Переглядів 1138 місяців тому
Neural network driven applications like ChatGPT suffer from hallucinations where they confidently provide inaccurate information. A fundamental reason for this inaccuracy is the lack of robust measures that are applied on the underlying neural network predictions. In this tutorial, we identify and expound on three human-centric robustness measures, namely explainability, uncertainty, and interv...
Quantization noise: an engineering approach
Переглядів 669 місяців тому
Presented by Prof. Zsolt Kollár This talk is a part of CSIP seminars at Georgia Institute of Technology. Abstract: Quantization is the process of approximating a continuous signal through a set of discrete values. While physical quantities are precise, when digitized and inserted into a computer or a digital system, they are rounded to the nearest representative value. This process introduces q...
Digital Image Processing - Image Denoising
Переглядів 237Рік тому
Digital Image Processing - Image Denoising using Wiener Filter, NLM, Transform Methods, BM3D, BLS-GSM, and Adaptive Curvelets
Digital Image Processing - Image Denoising
Переглядів 463Рік тому
Digital Image Processing - Image Denoising using spatial methods including rank filters, median filters, and spatially adaptive filters www.ghassanalregib.info
Attaining Sparsity in Large Language Models: Is It Easy or Hard?
Переглядів 89Рік тому
Presented by Dr.Atlas Wang, Associate Professor @ University of Texas at Austin This talk is a part of CSIP seminar series at Georgia Institute of Technology. Abstract: In the realm of contemporary deep learning, large pre-trained transformers have seized the spotlight. Understanding the underlying frugal structures within these burgeoning models has become imperative. Although the tools of spa...
Enabling Consistent Data Selection with Representation Shifts
Переглядів 33Рік тому
Presented by Ryan Benkert, PhD Candidate This talk is a part of CSIP seminar series at Georgia Institute of Technology. Abstract: Regression describes the performance deterioration after a model update. For modern data acquisition pipelines, performance regression is a major concern as models are updated iteratively with newly acquired data. However, the current standard in several data selecti...
A broad perspective on ASR research and quality at Google
Переглядів 46Рік тому
Presented by Dr. Parisa Haghani, Google Speech
Beyond UCB: The curious case of non-linear ridge bandits
Переглядів 106Рік тому
Presented by Nidev Rajaraman, PhD This talk is a part of CSIP seminar at Georgia Institute of Technology. Abstract: There is a large volume of work on bandits and reinforcement learning when the reward/value function satisfies some form of linearity. But what happens if the reward is non-linear? Two curious phenomena arise for non-linear bandits: first, in addition to the "learning phase" with ...
New advances in the decomposition and analysis of nonstationary signals
Переглядів 636Рік тому
Presented by Professor Antonio Cicone This talk is a part of CSIP seminars at Georgia institute of Technology. Abstract: In many applied fields of research, like Geophysics, Medicine, Engineering, Economy, and Finance, to mention a few, classical challenging problems are the identification of hidden information and features contained in a given signal, like quasi-periodicities and frequency pat...
Customizing Federated Learning to the Edge
Переглядів 99Рік тому
Presented by Professor Venkatesh Saligrama , Boston University, Amazon Scholar This talk is a part of CSIP seminars at Georgia Institute of Technology. Abstract: We propose a novel method for federated learning that is customized to the objective of a given edge device. In our proposed method, a server trains a global meta-model by collaborating with devices without actually sharing data. The t...
Two-Agent Competitive Reinforcement Learning
Переглядів 341Рік тому
Presented by Sihan Zeng, PhD Abstract: Multi-agent reinforcement learning studies the sequential decision making problem in the scenario where multiple agents co-exist in the same environment and jointly determine the environment transition and/or reward function. In this talk we consider two specific multi-agent settings and discuss the structure of the underlying optimization problems. The fi...
HandsOff: Labeled Dataset Generation with No Additional Human Annotations
Переглядів 136Рік тому
Presented by Austin Xu, PhD This talk is a part of CSIP seminars at Georgia Institute of Technology. Abstract: Recent work leverages the expressive power of generative adversarial networks (GANs) to generate labeled synthetic datasets. These dataset generation methods often require new annotations of synthetic images, which forces practitioners to seek out annotators, curate a set of synthetic ...
Representing the human brain data using networks that change with time
Переглядів 298Рік тому
Presented by Dr. Ashkan Faghiri This talk is a part of CSIP seminars at Georgia Institute of Technology. Abstract: The human brain is a complex and dynamic system, and our many measurements of this organ can be used to capture these two properties. We often need to perform extensive processing on these measurements before we can analyze and discuss the complexity and dynamism of the human brain...
Audio Classification with Small Training Datasets
Переглядів 417Рік тому
Presented by Alexander Lerch, Associate Professor and Director of Graduate Studies at the School of Music, Georgia Institute of Technology The talk is part of CSIP seminars at Georgia Institute of Technology. Abstract: Many tasks in music and audio classification lack large datasets and researchers thus struggle to train deep state-of-the-art networks with a large number of hyperparameters. Thi...
Enhancing Geophysics Data Interpretation through a Human-in-the-Loop Framework
Переглядів 189Рік тому
Enhancing Geophysics Data Interpretation through a Human-in-the-Loop Framework
[ICIP2021] Explaining Deep Models Through Forgettable Learning Dynamics
Переглядів 513 роки тому
[ICIP2021] Explaining Deep Models Through Forgettable Learning Dynamics
[ICIP2021] Open-Set Recognition with Gradient-Based Representations
Переглядів 3063 роки тому
[ICIP2021] Open-Set Recognition with Gradient-Based Representations
[ICIP2021] Extracting Causal Visual Features for Limited Label Classification
Переглядів 753 роки тому
[ICIP2021] Extracting Causal Visual Features for Limited Label Classification
[ICIP 2021] MAN-RECON: Manifold Learning for Smart Seismic Interpretation
Переглядів 843 роки тому
[ICIP 2021] MAN-RECON: Manifold Learning for Smart Seismic Interpretation
KLT - Karhunen-Loève Transform - KLT Transform of images
Переглядів 2,4 тис.4 роки тому
KLT - Karhunen-Loève Transform - KLT Transform of images
Discrete Wavelet Transform of Images (Haar and Hadamard)
Переглядів 9 тис.4 роки тому
Discrete Wavelet Transform of Images (Haar and Hadamard)
Discrete Cosine Transform (DCT) and Discrete Sine Transform (DST) of Images
Переглядів 1,6 тис.4 роки тому
Discrete Cosine Transform (DCT) and Discrete Sine Transform (DST) of Images
Discrete Fourier Transform (DFT) and Discrete Cosine Transform (DCT)
Переглядів 2,2 тис.4 роки тому
Discrete Fourier Transform (DFT) and Discrete Cosine Transform (DCT)
Spatial Interpolation in Image Processing
Переглядів 5644 роки тому
Spatial Interpolation in Image Processing
[AAPG2020] Weakly-Supervised Interpretation using Projection Matrices for Latent Space Factorization
Переглядів 714 роки тому
[AAPG2020] Weakly-Supervised Interpretation using Projection Matrices for Latent Space Factorization
Digital Image Processing: 2D Sampling
Переглядів 1,3 тис.4 роки тому
Digital Image Processing: 2D Sampling
[ICIP2020] Self-Supervised Annotation of Seismic Images using Latent Space Factorization
Переглядів 1314 роки тому
[ICIP2020] Self-Supervised Annotation of Seismic Images using Latent Space Factorization
Wow! Impressive simulation. Please what software did you use to create this simulated video?
Promo`SM 😱
great presentation. Would be even better if you add some links to public papers? Thanks.
May I ask if the latest video about ISAC will be released? I really want to see it. Thanks
May I ask if the latest video about ISAC will be released? I haven't seen it in time. Thanks.
Unfortunately we don't have the video of that talk.
❤
好!
Hi Prof. Ghassan AlRegib, great lectures, it would be great to share the slides.
Thanks a lot!
Excellent lecture!
Thank you for your explanation of this topic! This helped me a ton.
Great lecture! Is there a part 2 of this?
Could you please share the source code
nice video could you share the code
Very clear, direct, and useful. Thank you
There is audio issue :(
very interesting video. thank you
Thank you so much Prof Ghassan.
great work
Your video is very useful, thank you!!!
Great explanation on contrast variable!!
Yes
You show that you can use an air mouse but not what we can do with it or how.
very impressive. How did you get lego cubes to highlight? I just clone the material, change some properties in the beginning and reassign the material pointer during run time. I am not sure if it is a good solution or not...