CANDLE: Computer Architecture and Machine Learning Lab
Welcome to the webpage of Computer Architecture aND machine LEarning (CANDLE) research lab. Dr Sparsh Mittal is now with ECE department (joint appointment with Mehta Family School of DS/AI) at IIT Roorkee and some of his students are at CSE, IIT Hyderabad!
We work on the intersection of many interesting areas, e.g., autonomous driving vehicles, computer architecture, processor architectures for machine learning, neural network accelerators, VLSI, high-performance computing, approximate computing, etc.
Sticky: This presentation shows 1-slide summary of the major works we have done in last 3-4 years. It is useful for (1) those interested in collaborating with our group and (2) student interested in doing internship in our research lab.
Sticky: We have released source-code of many techniques proposed in our lab. See this github link.
2024/03: Dr Sparsh has been invited to serve on the TPC of "HPC Architecture" track of HiPC 2024 conference, "Architecture and Network" Track of SC 2024 conference, and as the track chair of "Hardware/Software for AI, Robotics, and Automation (AIR)" in the iSES 2024 conference.
2024/03: "Mott Memristor Based Stochastic Neurons for Probabilistic Computing", co-authored by Aabid, Dr Sparsh and Dr Farooq got accepted in IOP Nanotechnology (IF 3.5).
2024/03: A consultancy project "Chakbandi.AI portal development for Uttar Pradesh Government" has been awarded to Skynet and IIT Roorkee. In IIT Roorkee, Prof Siddhartha Khare from Civil Engg department is the PI. Dr Sparsh, along with two more professors are the Co-PIs.
2024/03: Maruthi defends his PhD from IIT Hyderabad and submits the thesis. Congratulations Dr Maruthi. He was co-advised by Dr Arjun Reddy (IITH) and Dr Sparsh.
2024/02: Our paper "Rethinking Intermediate Layers design in Knowledge Distillation for Kidney and Liver Tumor Segmentation" got accepted in ISBI (International Symposium on Biomedical Imaging) 2024, to be held in Greece.
2024/01: Our paper "CLINet: A Novel Deep Learning Network for ECG Signal Classification" got accepted in Journal of Electrocardiology.
2024/01: Our paper "Security Implications of Approximation: A Study of Trojan Attacks on Approximate Adders and Multipliers" received the best student paper award at VLSID 2024 (Cash prize of 30K INR).
2024/01: Our paper "TEFAR: An Efficient Transparent Finer-grained Encryption of Internet Access Artifacts" received the best paper award at COMSNETS India Internet Governance Workshop 2024.
2023/12: Prof Sparsh delivered an invited talk in the webinar on "Artificial Intelligence for Medicine" organized by AIIMS Rishikesh under the aegis of Telemedicine Society of India (Uttarakhand Chapter).
2023/12: Out of 2042 submitted papers, WACV 2024 accepted 847 papers, of which 53 were accepted as orals. One of our paper "Textual Alchemy: CoFormer for Scene Text Understanding" has been accepted as a oral paper in WACV 2024!
2023/10: "RAxC: Reflexivity-based Approximate Computing techniques for efficient remote sensing" got accepted in IEEE International Conference on Big Data, Italy.
2023/10: Three papers got accepted in WACV, a CORE-A rank conference in computer vision.
SynergyNet: Bridging the Gap between Discrete and Continuous Representations for Precise Medical Image Segmentation
Textual Alchemy: CoFormer for Scene Text Understanding
LIVENet: A novel network for real-world low-light image denoising and enhancement
2023/10: Vishesh Mishra, a PhD student at CSE department, IIT Kanpur has got PMRF fellowship. Vishesh has been doing internship/collaboration in/with CANDLE lab for last 2 years and has published 5 papers with Dr Sparsh.
2023/10: "ZETA: A Zero-Trust Security based Forensic-Ready Solution for Perimeter-less Enterprise Networks" got accepted in IEEE ANTS 2023.
2023/10: " Security Implications of Approximation: A Study of Trojan Attacks on Approximate Adders and Multipliers" got accepted in IEEE VLSID 2024.
2023/10: "At the Locus of Performance: Quantifying the Effects of Copious 3D-Stacked Cache on HPC Workloads" got accepted in ACM Transactions on Architecture and Code Optimization (TACO) (impact factor 1.6).
2023/10: "PaCL: Patient-aware Contrastive Learning Through Metadata Refinement for Generalized Early Disease Diagnosis" got accepted in Computers in Biology and Medicine (impact factor of 7.7).
2023/09: Dr Sparsh is invited to be TPC of 24th IEEE/ACM International Symposium on Cluster, Cloud, and Internet Computing (CCGrid) 2024, taking place in Philadelphia, USA. Dr Sparsh is also serving as a reviewer of WACV 2024.
2023/09: Dr Sparsh is conducting an short-course on RISC-V Assembly Language Programming. The PPTs and video recordings of this course are here.
2023/09: "LiBERTy: A Novel Model for Natural Language Understanding" got accepted in AIMLSystems 2023 (Bengaluru). This paper got the "best paper award honorable mention".
2023/09: " A Survey of Techniques for Optimizing Transformer Inference" got accepted in Journal of Systems Architecture (IF 4.5).
2023/07: Yash Jain and Vishu Saxena (BTech students), advised by Dr Sparsh for BTech project, have won the "Best Innovation" award in the “PROFESSIONAL DEVELOPMENT AND INNOVATION AWARD” scheme at IIT Roorkee. This award was established by 1983 batch of alumni of IIT Roorkee.
2023/07: M. Krishan Chaitanya, whom I advised for MTech thesis, has won the award under the category of Judges' Choice, in the “PROFESSIONAL DEVELOPMENT AND INNOVATION AWARD” scheme.
2023/07: Our survey paper on optimizing transformer inference is out.
2023/07: Yash Jain and Vishu Saxena, advised by Dr Sparsh for BTech project, have won the best BTech project in ECE Department.
2023/07: "VADF: Versatile Approximate Data Formats for Energy-Efficient Computing" got accepted in CASES, to be published in ACM Transactions on Embedded Computing Systems (TECS).
2023/05: "Machine Learning and Polynomial Chaos models for Accurate Prediction of SET Pulse Current" got accepted in ISVLSI 2023.
2023/05: Ananya Mantravadi (an intern in our lab from IIIT Raichur) got admit for MCS at NCSU, USA. She has published one IEEE conference paper during internship. Congratulations.
2023/04: Our paper on text erasure got accepted in ICDAR 2023 (USA). ICDAR is Core A rank conference.
2023/03: Kumud Lakara (a former intern in our lab from Manipal Institute of Technology) got admit in MS at University of Oxford. She had published 2 IEEE conference papers during internship. Congratulations. Srishti Srivastava got admission in MS progam at Columbia University last year.
2023/03: Phani Jayanth (a former intern in our lab from IIT Madras) got admit in MS at ETH Zurich. He had published a paper in IEEE TNNLS during internship. Congratulations.
2023/03: Dr Sparsh delivers an invited talk at Qualcomm Tech Talk Series.
2023/02: "PCBSegClassNet - A Light-weight Network for Segmentation and Classification of PCB Component" accepted in Expert Systems with Applications (impact factor 8.665). "
2023/02: " An Active Memristor based Rate-Coded Spiking Neural Network" accepted in Neurocomputing (Impact factor 5.8).
2023/01: Dr Sparsh is invited as a panelist for the fireside chat on "HPC and AI: Two sides of the same coin." Sincere thanks to ACM India and TCS Research for extending the invitation.
2023/01: With UG interns in India, we have published 26 papers in last 6+ years. Kudos to those interns and sincere thanks to the mentors and other co-authors on those papers. See the full list of interns and papers here. Especially, Dr Sai Chandra Teja has mentored many interns on multiple papers.
2023/01: Dr Sparsh is invited to serve on TPC of SC'23 (Machine learning and HPC track). SC'23 is world's largest conference on HPC with 10K+ attendees.
2023/01: 3 papers accepted in ISQED'23 (California, USA).
2022/12: "TPFNet: A Novel Text In-painting Transformer for Text Removal" is accepted in a workshop at NeurIPS-22. The following two sample videos show the results of TPFNet network. In these videos, the left-most frame shows the original video. The middle frame shows the video after text-erasure by TPFNet. The right-most frame shows the attention map of text.
Poem video: https://youtu.be/nkYF0GlZgi8
Sports news video: https://youtu.be/KAPSqhlz_6Y
2022/11: SERB Core Research Grant Proposal on "Designing Secure and Robust Artificial Intelligence (AI) Algorithms and Accelerators" is approved.
2022/11: "Realizing CNN Inference in Log-domain" is accepted in a workshop at AAAI-23.
2022/10: "A Machine Learning-based model for Single Event Upset Current Prediction in 14nm FinFETs" accepted in VLSID'23. Congratulations to Vibhu and Vivek Kumar.
2022/09: A paper "Ensembling Deep Learning And CIELAB Color Space Model for Fire Detection from UAV images" is accepted in AIMLSystems 2022.
2022/08: A paper, "GAFNet: A Global Fourier Self Attention Based Novel Network for multi-modal downstream tasks " is accepted in IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) 2023 Hawaii, USA.
2022/08: "ClarifyNet: A High-Pass and Low-Pass Filtering Based CNN for Single Image Dehazing " has been accepted in Journal of Systems Architecture (IF 5.86). The source-code of ClarifyNet is here.
2022/08: Subhrajit defends his PhD thesis. Congratulations Dr Nag.
2022/07: A paper is accepted in Physics in Medicine and Biology (Impact factor 4.174)
2022/06: A paper is accepted in Taylor & Francis "Remote Sensing Letters" (IF 2.58).
2022/06: Dr Sparsh is serving as TPC member of ICPP 2022 and NAS 2022. With Maruthi and Dr Arjun, we submitted an Indian patent application on "METHOD AND SYSTEM FOR PRIVACY-PRESERVING CONTINUOUS INTERNET FORENSICS".
2022/06: "WaferSegClassNet - A Light-weight Network for Classification and Segmentation of Semiconductor Wafer Defects" is accepted in Computers in Industry journal (Impact factor 7.63).
2022/06: A paper got accepted IEEE TED. We have released four medical datasets: COVID/pneumonia chest X-ray images, fundus images dataset, and two datasets on breast mammography images (1 and 2).
2022/04: ECE, IITR is offering a PG Certification in VLSI on Coursera. In this program, Dr Sparsh is offering a course "Emerging Technologies: Accelerators for AI".
2022/04: Dr Sparsh gave a training session on "Computer Architecture for AI" to Qualcomm engineers. Sincere thanks to Sushim Shrivastava and Madhu Kandasamy for inviting.
2022/04: Our collaborative work with RIKEN, Tokyo Institute of Technology and others got covered in The Next Platform. In this paper, we show that use of a 3d-stacked cache can provide 10X speedup for the A64FX supercomputer by the year 2028.
2022/04: Two papers accepted in 5th IEEE Conference on Dependable and Secure Computing (DSC) (Edinburgh, UK).
2022/04: "MEGA-MAC: A Merged Accumulation based Approximate MAC Unit for Error Resilient Applications" accepted as a short research paper in GLSVLSI 2022 (USA).
2022/01: Dr Sparsh gave classes in Computer Architecture Winter School (CAWS) organized by National Supercomputing Mission (NSM). Thanks to Dr Rupesh, Dr Madhura and Dr Vishwesh for inviting me and organizing everything.
2022/01: Maruthi's paper accepted in IEEE Access (IF 3.36). Our survey on underwater image classification accepted in IEEE TNNLS (IF 10.45)
2022/01: The link to video lectures of "computer architecture for AI" course (taught at IIT Roorkee in Aug-Nov 2021 semester) are available here. This course is very useful for researchers working at intersection of AI and computer architecture.
2021/11: Dr Sparsh is offering one executive course on accelerators for AI and one on Computer Architecture for AI.
2021/10: In the Stanford's 2021 list of world's top researchers, in the field of Computer Hardware & Architecture, Dr Sparsh is ranked as number 71 (for whole career) and as number 3 (for year 2020 alone).
2021/10: A collaborative work with Dr Brajesh K. Kaushik, titled, "High-performance Voltage Controlled Multilevel MRAM Cell" is accepted by IOP SST (IF 2.5). Congratulations to Arshid, Seema and Dr Kaushik.
2021/10: IITR establishes a School of Data Science and Artificial Intelligence. Dr Sparsh was the member of core-team that prepared the presentation/proposal for this and presented it initially to Mehta Family foundation.
2021/10: A paper on DNN-based crowd-counting from drone images got accepted in IEEE Indicon 2021. Pre-trained model is here. Dr Sparsh's citations on Google Scholar has crossed 5000.
2021/09: Two papers got accepted in ACM AI-ML-Systems. The PMC dataset for reverse-engineering DNN is available here.
2021/08/30: Same day, two papers got accepted: one in ACM TECS (IF 1.9) and one in Elsevier JSA (IF 3.77)!
Kumud Lakara (UG intern in our lab, studying at Manipal Institute of Technology) was selected as one of the best interns by the SPARK committee (IIT Roorkee) in year 2021. 9 interns from IITR were selected as the best and Kumud is one of them.
2021/08: At IITR, Dr Sparsh is (or has been) a member of following committees:
IITR: institute academic ethic committee
ECE: TA allocation, DAPC, HPC purchase, faculty search committee
CAIDS: CAPC, CAC
ICC: faculty-advisory committee and purchase committee
2021/07: Two papers accepted in IEEE TED (IF 2.91) and one in Physica B: Physics of Condensed Matter (IF 1.9).
2021/06: Dr Sparsh is serving as TPC member of IPDPS 2022.
2021/06: A Survey of Hardware Architectures for Generative Adversarial Networks got accepted in JSA (IF 2.55).
2021/05: A Survey on Hardware Security of DNN Models and Accelerators got accepted in JSA (IF 2.55).
2021/05: Survey paper on SRAM-based processing-in-memory is out.
2021/04: Dr Sparsh joins the advisory board of DeepEigen, an online-education startup. Dr Sparsh played an important role in launching the Center of AI & DS at IIT Roorkee and the Online MTech in VLSI program at ECE, IIT Roorkee in year 2021.
2021/04: Maruthi got the cloud innovator of the year award. Congratulations.
2021/04: Under IITR's SPARK internship for international students, two students: Yuval (UNSW, Australia) and Sarvasv (McGill University, Canada) have joined our lab. Welcome.
2021/04: A Survey of Deep Learning on CPUs: Opportunities and Co-optimizations got accepted in IEEE Transactions on Neural Networks and Learning Systems (TNNLS) (IF 8.79)! It got covered by HPCWire and TechDecoded.intel.io
2021/04: "A Survey of Deep Learning Techniques for Vehicle Detection from UAV Images" is accepted in Journal of Systems Architecture (IF 2.55).
2021/02: Placements of our recent UG interns: Sarthak Narayan admitted to Masters at Univ of Toronto; Kakul into MTech CS@IITD, Shreyas into MS at UT Austin, Srishti got job in PolyMage, Himanshi in Micron, Sai Harsh Kottapalli in Microsoft Azure cloud. Kudos to Maruthi for mentoring Sai Harsh and Kakul, and Poonam for mentoring Sarthak and Nandan for mentoring Shreyas.
2021/01: Two papers are accepted in Journal of Systems Architecture: (1) "CURATING: A Multi-Objective based Pruning Technique for CNNs" and (2) "A Survey of Accelerator Architectures for 3D Convolution Neural Networks"
2020/11: In the Stanford's list of world's top researchers, in the field of Computer Hardware & Architecture, Dr Sparsh is ranked as number 107 (for whole career) and as number 3 (for year 2019 alone).
2020/10: Dr Sparsh is elevated to IEEE senior member position.
2020/10: Our research on mobile-phone usage detection was praised by Education Minister of India (1, 2, 3). It was covered by TOI, HinduBusinessLine, TelanganaToday, Andhrajyothi and many other media platforms.
2020/08: Survey paper on Techniques for Intermittent Computing has been accepted in Journal of Systems Architecture.
2020/08: Our paper "Modeling Data Reuse in Deep Neural Networks by Taking Data-Types into Cognizance" got accepted into IEEE Transactions on Computers (Special Issue on Machine-Learning Architectures and Accelerators) (impact factor 2.71).
2020/07: Our paper "DeepPeep: Exploiting Design Ramifications to Decipher the Architecture of Compact DNNs" got accepted in ACM Journal on Emerging Technologies in Computing Systems (JETC) .
2020/07: Survey paper on "Hardware Accelerators and Optimization Techniques for RNNs" accepted in Journal of System Architecture. It got covered by HPCWire
2020/07: Our paper "Improving Accuracy and Efficiency of Object Detection Algorithms using Multiscale Feature Aggregation Plugins" is accepted in ANNPR 2020 (Switzerland).
2020.07: The paper "Detecting Usage of Mobile Phones using Deep Learning Technique" is accepted in EAI GOODTECHS 2020. The pretrained model and dataset are here.
2020/05: Nandan gets admission into PhD at New York University. Congratulations. Four MTech students are graduating this semester: Nandan, Poonam, Suraj and Santanu.
2020/05: Our paper on "DRACO: Co-Optimizing Hardware Utilization, and Performance of DNNs on Systolic Accelerator" is accepted in ISVLSI 2020.
2020/04: Dr Sparsh's talks on writing effective research papers and making presentations are here.
2020/03: Dr Sparsh's article on Brain-Drain is published in Eduvoice.
2020/03: Survey paper on Xeon Phi is accepted in CPE 2020. It is covered by InsideHPC.
2019/12: Dr Sparsh Mittal moved to ECE, IIT Roorkee. Suraj is placed in Microsoft and Poonam in Adobe.
2019/12: Paper on attention module for compact CNNs accepted in IEEE WACV 2020.
2019/11: Survey paper on reliability of DNN algorithms and architectures accepted in JSA.
2019/09: Paper on group convolution accepted in VLSID 2020.
2019/08: Survey paper on Deep Learning on GPU got accepted in JSA. It got covered by HPCWire.
2019/08: Bhargav is co-guided by Dr Biswa from IIT Kanpur and he will spend one semester (Aug-Dec 2019) at IIT Kanpur.
2019/07: Our survey paper on Micron's automata processor got accepted in JSA. It got covered by InsideHPC.
2019/07: Dr Sparsh delivered an invited talk at Xilinx Hyderabad. Thanks to Mr Kishore for hosting!
2019/06: Our survey paper on Xeon Phi is here!
2019/04: Shraiysh Vaishay has been selected for summer internship at NTT-AT, Japan and Chander Shekhar has been selected for internship at University of Tokyo, Japan. Congratulations to them.
2019/04: Our paper on Approximate Value Prediction in GPUs accepted in ICS 2019. This is a collaborative work with Dr Adwait Jog (College of William and Mary, USA).
2019/03: Poonam has been selected for summer internship at Fujitsu, Japan. Congratulations to Poonam!
2019/03: Subhrajit Nag has been selected for IITH-Swinburne University of Technology (Australia) joint PhD program. Congratulations to Subhrajit!
2019/01: Survey Paper on Deep Learning on NVIDIA Jetson Platform is accepted in JSA!
2018/12: Survey paper on mobile web browsing accepted in Concurrency and Computation.
2018/11: Survey paper on Spintronic architectures for PIM and neural networks accepted in JSA.
2018/11: Our recent survey papers have been covered by TheMemoryGuy website. Survey paper on Encoding techniques for reducing data-movement energy accepted in JSA.
2018/11: Our VLSID 2019 paper is one of the 12 (out ot 84 total) papers to get invitation to submit its extension for a special issue of IEEE TVLSI.
2018/10: Survey paper on FPGA-based accelerators for CNNs accepted in NCAA. It got covered by InsideHPC website.
2018/09: Survey paper on DRAM reliability accepted in Elsevier JSA. It got covered by StorageSearch website.
2018/09: Paper on ramifications of making DNNs compact is accepted in VLSID 2019.
2018/08: Received funding from Semiconductor research corporation (SRC) for research on autonomous driving vehicles.
2018/06: Happy to host Mr. Sanjeev from Swaayatt Robots for a talk on "self-driving technology in India".
2018/05: Irfan is selected for summer internship in Japan under Sakura Science Plan. Congratulations to Irfan!
2018/05: Survey paper on GPU security techniques is accepted in HASS. Our article on GPU security is covered in HPCWire.
2018/05: Survey paper on branch predictor is highlighted on InsideHPC.com. Survey paper on ReRAM-based NN architectures is highlighted on StorageSearch.com.
2018/04: Survey paper on branch predictor and survey paper on memristor (ReRAM)-based PIM and neural network architectures accepted.
2018/04: We have an MTech (RA) position for Aug 2018 round of admission. Apply here.
2018/02: Survey paper on security-techniques for non-volatile memory is accepted in Springer HASS 2018.
2017/11: We have PhD position and MTech (RA) position for Dec 2017 round of admission. The PhD position is the prestigious "Intel PhD Fellowship". Deadline for application has passed.
2017/10: Dr Sparsh's proposal for recruiting student on Intel PhD Fellowship 2018 is approved
2017/10: Dr Sparsh gives seminar on "Improving Clarity in Paper-Writing". See the video recording here (3 minute onwards).
2017/04: Dr Sparsh receives ECR grant from SERB.
2017/01: We have formed CANDLE research lab. Stay tuned.