Slide 1: Pushing the boundaries of AI research
September 2, 2020 @qualcomm_tech Pushing the boundaries of A I research Qualcomm Technologies, Inc.
Slide 2: Our AI leadership
Our AI leadership Over a decade of cutting-edge AI R&D, speeding up commercialization and enabling scale Research face MWC demo showcasing Collaboration Brain Corp Qualcomm® Power efficiency Mobile AI Enablement Center detection with photo sorting and with Google on raises $114M Artificial Intellige
Slide 3: Advancing AI research to make efficient AI ubiquitous
Advancing AI Perception research to make Object detection, speech recognition, contextual fusion IoT/IIoT efficient AI ubiquitous Edge cloud Power efficiency Personalization Efficient learning Reasoning Model design, Continuous learning, Robust learning Scene understanding, language compression, qua
Slide 4: Leading research and development
Leading research and development Across the entire spectrum of AI Fundamental Applied research research Deep Neural Graph and Deep AI Model G-CNN transfer network kernel learning for Efficiency Toolkit learning compression optimization graphics Bayesian Deep Neural Compute Source CV DL for combinato
Slide 5: Leading research to efficiently quantize AI models
Trained neural network model New Inference input data output Quantization Automated reduction in precision of weights Promising results show that and activations while maintaining accuracy low-precision integer inference can become widespread Models trained at Inference at Virtually the same accurac
Slide 6: Pushing the limits of what's possible with quantization
Pushing the limits of what’s possible with quantization Data-free quantization AdaRound Bayesian bits How can we make quantization Is rounding to the nearest value the best Can we quantize layers to different bit widths based on as simple as possible? approach for quantization? precision sensitivity
Slide 7: Leading quantization research and fast commercialization
Leading quantization research and fast commercialization Driving the industry towards integer inference and power-efficient AI Quantization Relaxed Quantization Data-free Quantization AdaRound Bayesian Bits research (ICLR 2019) (ICCV 2019) (ICML 2020) (arXiv 2020) Quantization commercialization Qual
Slide 8: AIMET makes AI models small
AIMET makes AI models small Open-sourced GitHub project that includes state-of-the-art quantization and compression techniques from Qualcomm AI Research Trained AI Model Efficiency Toolkit Optimized AI model (AIMET) AI model Compression Quantization TensorFlow or PyTorch If interested, please join t
Slide 9: An increasing demand for energy efficient video processing
An increasing demand Developing AI approaches for energy efficient to reduce unnecessary video processing computation Valuable information is being extracted from video Recognizing redundancy between video frames streams across diverse devices and use cases so that the same thing is never computed t
Slide 10: Conditional computation using gated networks
Conditional computation using gated networks Introduce gates with low computation cost to skip unnecessary computation in neural networks Problem State-of-the-art accuracy while reducing A large portion of the neural network is computation by up to 3X not necessary for the prediction, wasting comput
Slide 11: Deep generative model research for unsupervised learning
Generative Powerful Broad models capabilities applications Deep generative Variational auto encoder (VAE)* Extract features by learning a low-dimension Speech/video compression model research feature representation for unsupervised Generative adversarial Sampling to generate, Text to speech learning
Slide 12: Deep generative model research for unsupervised learning
Input unlabeled Example use case: data, such as encoder/decoder images Generative models Encoder part Deep generative Variational auto encoder (VAE)* model research Extract features by learning a low-dimension Sampling to generate, restore, predict or for unsupervised Generative adversarial feature
Slide 13: Novel machine learning-based video codec research
Novel machine learning-based video codec research Neural network based I-frame and P-frame compression Deep End-to-end deep learning neural net Entropy bits Entropy T=1 I-frame encoding decoding I-frame Ground truth encoder decoder Warp Reconstruction Entropy bits Entropy T=2 Motion P-frame encoding
Slide 14: Applying AI to solve difficult wireless challenges
Applying AI to solve difficult wireless challenges Deep wireless domain knowledge is required to optimally use AI capabilities Wireless challenges AI strengths Hard-to-model problems Learning representations for hard-to-model problems Computational infeasibility Training policies of optimal solution
Slide 15: Applying AI to RF for centimeter-accurate positioning
Scattering RF splits or is an aggregate of many reflections due to complex surfaces Applying AI to RF for centimeter-accurate positioning Precise indoor positioning is valuable across industries and for location-aware services Diffraction The RF path is bent Modeling indoor RF propagation is complex
Slide 16: Neural unsupervised learning from RF for positioning
Neural unsupervised learning from RF for positioning Injecting domain knowledge for interpretability Real Tx Virtual Tx Virtual Tx locations (learnable) Real map Learned map 𝑝𝐴 𝑝1 𝑝2 𝑝3 (CSI vs position) Unsupervised Line of Reflected path is Neural Receiver sight also equivalent to network…
Slide 17: Teaching Cars to See
Teaching Cars to See \ +
Slide 18: AI Radar
AI RADAR Perceptive radar Research results Traditional radar is affordable and responsive, Significant improvements in position and size has a long range, measures velocity directly, and estimation, velocity estimation, object classification, isn’t compromised by lighting or weather conditions and u
Slide 19: AI radar detects occluded vehicles in complex scenarios
AI radar detects occluded vehicles in complex scenarios Video shows the accuracy between AI radar and ground truth up to 94 meters Performance in complex scenarios Robust performance even in crowded scenarios 20
Slide 20: From Bayesian bits to quantum bits
Superposition Entanglement Quantum computing Each qubit is both Qubits that are inextricably linked Utilize quantum mechanics to 1 and 0 at the same time such that whatever happens to achieve exponential speedup one immediately affects the other From Bayesian bits to quantum bits Through quantum com
Slide 21: Applying quantum mechanics to machine learning
Applying quantum mechanics to machine learning Fundamental green field research to utilize the exponential power of quantum computing on various use cases Quantum annealing Quantum deep learning Combinatorial optimization problems are widespread, The statistics of quantum mechanics can apply to deep
Slide 22: Quantum deformed binary neural networks
Quantum Measurement deformed binary Output Qubit (readout prob output) neural networks O1 H ● ● ● FT † Running a classical neural Ot H network on quantum computer Transform X1 (t-1) U1 U1 2 Key ideas W1 X2 2 (t-1) U2 U2 Deform a classical neural network into a W2 ●●● quantum neural network with quan
Slide 23: Advancing AI research to make power efficient AI ubiquitous
Advancing AI research to make power efficient AI ubiquitous — from device to cloud Conducting leading research and development across the entire spectrum of AI Creating an AI platform fundamental to scaling AI across the industry and applications 24
Slide 24: Questions?
www.qualcomm.com/ai www.qualcomm.com/news/onq @qualcomm_tech Questions? http://www.youtube.com/playlist?list=PL8AD95E4F585237C1&feature=plcp Connect with Us http://www.slideshare.net/qualcommwirelessevolution 25
Slide 25: Thank you
Thank you Follow us on: For more information, visit us at: www.qualcomm.com & www.qualcomm.com/blog Nothing in these materials is an offer to sell any of the References in this presentation to “Qualcomm” may mean components or devices referenced herein. Qualcomm Incorporated, Qualcomm Technologies,