Deep Learning Projects
From iisprojects
Contents
 1 What is Deep Learning?
 2 Foundation models and LLMs for Health
 3 Hardwareoriented neural architecture search (NAS)
 4 Algorithms & Frameworks for Quantization and Deployment for Deep Neural Networks (DNNs)
 5 Hardware Acceleration of DNNs and QNNs
 6 EventDriven Computing
 7 OnDevice Training
 8 Prerequisites
 9 Tags
 10 Available Projects
 11 Projects in Progress
 12 Completed Projects
What is Deep Learning?
Nowadays, machine learning systems are the goto choice when the cost of analytically deriving closedform expressions to solve a given problem is prohibitive (e.g., it is very timeconsuming, or the knowledge about the problem is insufficient). Machine learning systems can be particularly effective when the amount of data is large, since the statistics are expected to get more and more stable as the amount of data increases. Amongst machine learning systems, deep neural networks (DNNs) have established a reputation for their effectiveness and simplicity. To understand this success as compared to that of other machine learning systems, it is important to consider not only the accuracy performance of DNNs, but also their computational properties. The training algorithm (an iterative application of backpropagation and stochastic gradient descent) is linear in the data set size, making it more appealing in big data contexts than, for instance, support vector machines (SVMs). DNNs do not use branching instructions, making them predictable programs and allowing to design efficient access patterns for the memory hierarchies of the computing devices (exploiting spatial and temporal locality). DNNs are parallelizable, both at the neuron level and at the layer level. These predictability and parallelizability properties make DNNs an ideal fit for modern SIMD architectures and distributed computing systems.
The main drawback of these systems is their size: millions or even billions of parameters are a common feature of many topperforming DNNs, and a proportional amount of arithmetic operations must be performed to process each data sample. Hence, to reduce the pressure of DNNs on the underlying computing infrastructure, research in computational deep learning has focussed on two families of optimizations: topological and hardwareoriented.
Topological optimizations are concerned with network topologies (AKA network architectures) which are more efficient in terms of accuracyperparameter or accuracyperMAC (multiplyaccumulate operation). As a specific form of topological optimization, pruning strategies aim at maximizing the number of zerovalued operands (parameters and/or activations) in order to 1) take advantage of sparsity (for storing the model) and to 2) minimize the number of effective arithmetic operations (i.e., the operations not involving zerovalued operands, which must be actually executed). Hardwareoriented optimizations are instead concerned with replacing timeconsuming and energyhungry operations, such as evaluations of transcendent functions or floatingpoint MAC operations, with more efficient counterparts, such as piecewise linear activation functions (e.g., the ReLU) and integer MAC operations (as in quantized neural networks, QNNs).
Foundation models and LLMs for Health
Incorporating Foundation Models and Large Language Models (LLMs) within artificial intelligence is gaining significant traction, particularly due to their potential applications in the health sector. This project is dedicated to developing sophisticated methodologies for utilizing foundation models and LLMs in healthrelated applications, specifically analyzing electroencephalogram (EEG) brain signals.
In healthcare and biomedical research, implementing advanced computational models, notably Foundation Models and Large Language Models (LLMs), revolutionizes the understanding and interpretation of intricate biosignals. We stand at the vanguard of this revolutionary change, delving into the capabilities of these models for the analysis and interpretation of critical biosignals, including electroencephalograms (EEG) and electrocardiograms (ECG).
Foundation Models, encompassing a spectrum of robust, pretrained models, are transforming our ability to process and interpret large datasets. Initially trained on extensive and diverse datasets, these models are adaptable for specific tasks, offering remarkable accuracy and efficiency. This adaptability renders them particularly beneficial for biosignal analysis, where the intricacies of EEG and ECG data demand both precision and contextual understanding.
As a subset of Foundation Models, LLMs have demonstrated efficacy in processing and generating human language. At IIS, we are pioneering the application of LLMs in the domain of biosignal interpretation, extending beyond textual data. This entails training the models to interpret the 'language' of biosignals, translating complex patterns into actionable insights.
Our emphasis on EEG and ECG signals is motivated by these biosignals' profound insights into human health. EEGs, capturing brain activity, and ECGs, monitoring heart rhythms, are instrumental in diagnosing and managing various health conditions. By leveraging Foundation Models and LLMs, our objective is to refine diagnostic accuracy, predict health outcomes, and customize patient care.
IIS invites Master's students to immerse themselves in this pioneering area. Our projects offer avenues to engage with stateoftheart technologies, apply them to realworld health challenges, and contribute to shaping a future where healthcare is more predictive, preventive, and personalized. We encourage your participation in this exhilarating endeavor to redefine the confluence of healthcare and technology.
Links
Dr. Philipp Mayer

Hardwareoriented neural architecture search (NAS)
The problems of topology selection and pruning can be considered instances of the classical statistics problems of model selection and feature selection, respectively. In the scope of deep learning, model selection is also called neural architecture search (NAS). When designing a DNN topology, you have a large number of degrees of freedom at your disposal: number of layers, number of neurons for each layer, connectivity of each neuron, and so on; moreover, the number of choices for each degree of freedom is huge. These properties imply that the design space for a DNN can grow exponentially, making exhaustive searches prohibitive. Therefore, to increase the efficiency of the exploration, stochastic optimization tools are the preferred choice: evolutionary algorithms, reinforcement learning, gradientbased techniques or even random graph generation. An interesting feature of model selection is that specific constraints can be enforced on the search space so that desired properties are always respected. For instance, given a storage budget describing a hard limitation of the chosen computing platform, the network generation algorithm can be limited to propose topologies that do not exceed a given number of parameters. This capability of incorporating HW features as constraints on the search space make NAS algorithms very interesting in the context of generating HWfriendly DNNs.
Thorir Mar Ingolfsson

Cristian Cioflan

Victor Jung

Algorithms & Frameworks for Quantization and Deployment for Deep Neural Networks (DNNs)
The typical training algorithm for DNNs is an iterative application of the backpropagation algorithm (BP) and stochastic gradient descent (SGD). When the quantization is not “aggressive” (i.e., when the parameters and feature maps can be represented as integers with a precision of 8bits or more), many solutions are available either in specialized literature or in commercial software that can convert models pretrained with gradient descent to quantized counterparts (posttraining quantization). But when the precision is extremely reduced (i.e., 1bit or 2bits operands), these solutions can no longer be applied, and quantizationaware training algorithms are needed. The naive application of gradient descent (which in theory is not even correct) to train these QNNs yields major accuracy drops. Hence, it is likely that suitable training algorithms for QNNs require to replace the standard BP+SGD scheme, which is suitable for differentiable optimization, with search strategies that are more apt for discrete optimization.
Victor Jung

Cristian Cioflan

Georg Rutishauser

Philip Wiese

Hardware Acceleration of DNNs and QNNs
Deep Learning (DL) and Artificial Intelligence (AI) are quickly becoming dominant paradigms for all kinds of analytics, complementing or replacing traditional data science methods. Successful atscale deployment of these algorithms requires deploying them directly at the data source, i.e. in the IoT endnodes collecting data. However, due to the extreme constraints of these devices (in terms of power, memory footprint, area cost), performing full DL inference insitu in lowpower endnodes requires a breakthrough in computational performance and efficiency. It is widely known that the numerical representation typically used when developing DL algorithms (singleprecision floatingpoint) encodes a higher precision than what is actually required to achieve high qualityofresults in inference (Courbariaux et al. 2016); this fact can be exploited in the design of energyefficient hardware for DL. For example, by using ternary weights, which means all network weights are quantized to {1,0,1}, we can design the fundamental compute units in hardware without using an HWexpensive multiplication unit. Additionally, it allows us to store the weights much more compact onchip.
Angelo Garofalo

Georg Rutishauser

Moritz Scherer

Arpan Suravi Prasad

Gamze İslamoğlu

Philip Wiese

EventDriven Computing
With the increasing demand for "smart" algorithms on mobile and wearable devices, the energy cost of computing is becoming the bottleneck for battery lifetime. One approach to defuse this bottleneck is to reduce the compute activity on such devices  one of the most popular approaches uses sensor information to determine whether it is worth to run expensive computations or whether there is not enough activity in the environment. This approach is called eventdriven computing. Eventdriven architectures can be implemented for many applications  From pure sensing platforms to multicore systems for machine learning on the edge. At IIS, we cover most of these applications. Besides working with novel, stateoftheart sensors and sensing platforms to push the limits of lifetime of wearables and mobile devices, we also work with cuttingedge computing systems like Intel Loihi for Spiking Neural Networks to minimize the energy cost of machine intelligence.
Alfio Di Mauro

Moritz Scherer

Arpan Suravi Prasad

OnDevice Training
The fast development of the Internetof Things (IoT) comes with the growing need for smart endnode devices able to execute Deep Learning networks locally. Processing the data on device has many advantages, not only drastically reducing the latency and communication energy cost, but also taking one step towards autonomous IoT endnodes. Most of the current research efforts are focusing on inference, under the "trainthendeploy" paradigm. However, this results in a device unable to face reallife phenomena such as data distribution shifts or class increments. At IIS, we are actively researching new methods to tackle this significant challenge in the context of tightly memory constrained devices such as Microcontrollers (MCUs).
Cristian Cioflan

Viviane Potocnik

Victor Jung

Prerequisites
We have no strict, general requirements, as they are highly dependent on the exact project steps. The projects will be adapted to the skills and interests of the student(s)  just come talk to us! If you don't know about GPU programming or CNNs or ... just let us know and we can together determine what is a useful way to go  after all you are here to learn not only about project work but also to develop your technical skills.
Only hard requirements:
 Excitement for deep learning
 For HW Design projects: VLSI 1, VLSI 2 or equivalent
Tags
All our projects will be categorized into three categories. Therefore, look out for the following tags:
 Algorithmic  you will mainly make algorithmic evaluations using languages and frameworks like e.g. Python, Pytorch, Tensorflow and our inhouse frameworks like Quantlab, DORY, NEMO
 Embedded Coding  you will implement e.g. ccode for one of our microcontrollers
 HW Design  you will be designing HW including writing RTL, simulate, synthesize, and layout (backend) some HW
Available Projects
New projects are constantly being added, check back often! If you have any questions or would like to propose own ideas, do not hesitate to contact us!
 Approximate Matrix Multiplication based Hardware Accelerator to achieve the next 10x in Energy Efficiency: Full System Intregration
 Audio Visual Speech Separation and Recognition (1S/1M)
 Deep Learning Based Anomaly Detection in ECG Signals Using Foundation Models
 Design of a Highperformance Hybrid PTZ for Multimodal Vision Systems
 Design of a Low Power Smart Sensing Multimodal Vision Platform
 Exploratory Development of a Unified Foundational Model for Multi Biosignal Analysis
 ExtremeEdge Experience Replay for Keyword Spotting
 GPT on the edge
 Learning at the Edge with HardwareAware Algorithms
 MixedPrecision Neural Networks for BrainComputer Interface Applications
 Modular Distributed Data Collection Platform
 NeuroSoC RISCV Component (M/12S)
 Object Detection and Tracking on the Edge
 Predict eye movement through brain activity
 Pretraining Foundational Models for EEG Signal Analysis Using Open Source Large Scale Datasets
 Realtime Gaze Tracking on Siracusa
 Softmax for Transformers (M/12S)
 System Emulation for AR and VR devices
 Testbed Design for Selfsustainable IoT Sensors
 Towards Flexible and Printable Wearables
 Training and Deploying NextGeneration Quantized Neural Networks on Microcontrollers
 Visualization of Neural Architecture Search Spaces
 GPT on the edge
 Testbed Design for Selfsustainable IoT Sensors
 Towards Flexible and Printable Wearables
 Modular Distributed Data Collection Platform
 Object Detection and Tracking on the Edge
 Design of a Low Power Smart Sensing Multimodal Vision Platform
 Design of a Highperformance Hybrid PTZ for Multimodal Vision Systems
 Integration Of A Smart Vision System
 Spiking Neural Network for Autonomous Navigation
 EventDriven Convolutional Neural Network Modular Accelerator
 Level Crossing ADC For a Many Channels Neural Recording Interface
Projects in Progress
 3D Matrix Multiplication Unit for ITA (1S)
 Efficient TNN compression
 Investigation of Quantization Strategies for Retentive Networks (1S)
 On  Device Continual Learning for Seizure Detection on GAP9
 Physical Implementation of ITA (2S)
 Streaming Layer Normalization in ITA (M/12S)
 Ternary Neural Networks for Face Recognition
Completed Projects
 Audio Visual Speech Recognition (1S/1M)
 Audio Visual Speech Separation (1S/1M)
 Bandwidth Efficient NEureka
 Bridging QuantLab with LPDNN
 Continual Learning for Adaptive EEG Monitoring in Epileptic Seizure Detection
 Evaluating SoA PostTraining Quantization Algorithms
 Exploring NAS spaces with CBRED
 Exploring schedules for incremental and annealing quantization algorithms
 Feature Extraction and Architecture Clustering for Keyword Spotting (1S)
 Feature Extraction for Speech Recognition (1S)
 Flexfloat DL Training Framework
 Mapping Networks on Reconfigurable Binary Engine Accelerator
 Neural Architecture Search using Reinforcement Learning and Search Space Reduction
 Online Learning of User Features (1S)
 ResourceConstrained FewShot Learning for Keyword Spotting (1S)
 Transformer Deployment on Heterogeneous ManyCore Systems