For open positions, see Jobs.
About
The Machine Analysis of Data for Human Audition and Visualization (MADHAV) Lab is a research group in the Department of Electrical Engineering at the Indian Institute of Technology Kanpur set up and led by Prof. Vipul Arora.
Our group’s research interests lie at the intersection of the theory and application of machine learning - with a focus on applications in Audio and Physics. Our group focuses in the following areas:
- Human Machine Learning
- Efficient data labeling
- Data efficient learning and adaptation
- Generative Machine learning
- Normalizing flows, adversarial learning, diffusion models
- Sampling methods: Monte Carlo, application in Physics (lattice QCD and statistical physics)
- Audio signal processing
- Audio LLMs: textless speech tokenization and fast search
- Music analysis
- Speech recognition
- Other applications
- Image generation
- Sensor calibration
See our lab outcomes here
News
- Oct 2024: Paper ("FlowHash: Accelerating Audio Search with Balanced Hashing via Normalizing Flow") accepted in IEEE/ACM Trans. Audio, Speech, Lang. Processing. Congratulations Anup.
- Oct 2024: Paper accepted in IEEE/ACM Trans. Audio, Speech, Lang. Processing. Congratulations Sagar.
- Oct 2024: Patent Granted: "Machine-Learning Based (ML-Based) Audio Tokenizing System And Method For Indexing And Retrieving Audio Sequences".
- Jun 2024: Prof. Arora gave an invited talk on "Mode-Collapse in Generative Models" at TU Berlin.
- Jun 2024: Paper accepted in IEEE/ACM Trans. Audio, Speech, Lang. Processing. Congratulations Nagarathna and Thishyan.
- May 2024: Paper accepted in IEEE Trans. Neural Networks and Learning Sys. Congratulations Swapnil.
- May 2024: Prof. Arora is awarded the PK Kelkar Research Fellowship for three years.
- May 2024: Paper accepted in Scipost Physics. Congratulations Vikas.
- May 2024: Paper accepted in IEEE/ACM Trans. Audio, Speech, Lang. Processing. Congratulations Kavya.
- Apr 2024: Patent Granted: "System and Method for Music Training".
- Mar 2024: Prof. Arora gave a talk on "Human machine learning for Searching Big Audio Data" at NIT Calicut.
- Feb 2024: Prof. Arora gave a keynote talk on "Human-machine learning" at ICMLA conference at Chitkara University.
- Dec 2023: WiSSAP 2023 school (AI for Music) concluded successfully. Videos and slides are here.
- Dec 2023: One paper on acoustic event detection using ontology-informed constraints accepted at the ICASSP 2024. Congratulations Akshay and Sayeedul!
- August 2023: We are organizing the 16th Winter School on Speech and Audio Processing 2023 (AI for Music) at IIT Kanpur. Register now
- July 2023: A new research project started with Monotype LLC. Positions open for students.
- June 2023: Patent Granted on A Sensor System to read MIDI from Indian Harmoniums. Congratulations Suraj!
- May 2023: One paper on acoustic word embeddings accepted at the INTERSPEECH' 23. Congratulations Adhiraj!
- Feb 2023: Three papers accepted at the ICASSP' 23. Congratulations Sumit, Anup and Akshay!
- Jan 2023: One paper on normalizing flows for lattice field theory accepted at the Physical Review D journal. Congratulations Ankur!
- Jan 2023: One paper on audio representation learning accepted at the ICLR' 23. Congratulations Adhiraj!
- Jan 2023: Kavya and Prof. Arora delivered a tutorial on Meta Learning at the CODS-COMAD' 23 at IIT Bombay. Video
- Dec 2022: One paper on generative learning for lattice Gross-Neveu model accepted at the Scipost Physics Core journal. Congratulations Ankur!
- Oct 2022: Prof. Arora delivered a talk on Model Adaptation for Learning from Small Data at C4DM, Queen Mary University of London.
- July 2022: One paper on Audio Fingerprinting accepted at the ISMIR' 22. Congratulations Anup!
