Deep Convolutional Neural Networks and future

Talk given at Bay Area Women in Machine Learning and Data Science (https://www.meetup.com/Bay-Area-Women-in-Machine-Learning-and-Data-Science/events/274038207/)

Convolutional Neural Networks, works on paradigm of sharing of weights in AI thereby creating kernels that filter information from the data. This simple idea along with open-source frameworks lead to many astounding work in DL community, along with implications in many business products that we use daily. We will understand CNNs as the basic level and try to brainstorm what holds for this architectural model in the future given new models and methods to train neural networks.

Transcript

1. Deep Convolutional Neural Networks TANISHA R BHAYANI AI RESEARCHER AND DEVELOPER TANISHA BHAYANI 1
2. Industrial Applications 2 TANISHA BHAYANI
3. Industrial Applications 3 TANISHA BHAYANI
4. Industrial Applications 4 TANISHA BHAYANI
5. The Convolution Operation Intuitively, the convolution of two functions represents the amount of overlap between the two functions. The function g is the input, f the kernel of the convolution. 5 TANISHA BHAYANI
6. The Convolution Operation 6 TANISHA BHAYANI
7. The Convolution Operation 7 TANISHA BHAYANI
8. The Convolution Operation 8 TANISHA BHAYANI
9. Convolutional Neural Networks A convolutional neural network (CNN) is a type of artificial neural network used in image recognition and processing that is specifically designed to process pixel data. CNNs are powerful image processing, artificial intelligence (AI) that use deep learning to perform both generative and descriptive tasks, often using machine vison that includes image and video recognition, along with recommender systems and natural language processing (NLP). 9 TANISHA BHAYANI
10. Convolutional Neural Network • Depthwise convolution is the channel-wise n×n spatial convolution. Suppose in the figure above, we have 5 channels, then we will have 5 n×n spatial convolution. • Pointwise convolution actually is the 1×1 convolution to change the dimension. 10 TANISHA BHAYANI
11. “When we’re learning to see, nobody’s telling us what the right answers are — we just look. Every so often, your mother says “that’s a dog”, but that’s very little information. You’d be lucky if you got a few bits of information — even one bit per second — that way. The brain’s visual system has 10¹⁴ neural connections. And you only live for 10⁹ seconds. So it’s no use learning one bit per second. You need more like 10⁵ bits per second. And there’s only one place you can get that much information: from the input itself.” — Geoffrey Hinton, 1996 11 TANISHA BHAYANI
12. Misclassification in CNN Similar Feature Set causes Misclassification 12 TANISHA BHAYANI
13. Misclassification in CNN Adversarial Attacks 13 TANISHA BHAYANI
14. Misclassification in CNN Perturbation in Real World dataset 14 TANISHA BHAYANI
15. Recent Research in DCNN 1. Vernacular OCR 2. Person Re-identification using Siamese Networks 3. Animal Facial Recognition using CNN 4. Accurate Recommendations Engine 5. 4D CNN for medical domain (Brain Imaging) 6. Image Generation de-noising using CNN 7. Clustering Environmental images using CNN, and more... 15 TANISHA BHAYANI
16. Recent Research in DCNN 1. Attention based models 2. Capsule Networks 3. Reinforcement Learning 4. GANS 16 TANISHA BHAYANI
17. Questions ? TANISHA R BHAYANI AI RESEARCHER AND DEVELOPER 17 TANISHA BHAYANI