"Building Vector-Symbolic Architectures with Python"
summary
Introduces hdlib, a Python library (GitHub and JOSS doi provided) that implements binary/bipolar hyperdimensional vectors, elementwise bind/permute/sum, and a retraining classifier that sums sample encodings and corrects class vectors on misclassification; a stepwise backward feature-elimination wrapper is included and a Galaxy tool (shopen2) integrates the model on usgalaxy.org. Code is on PyPI/conda with ~3,000 downloads, the repo is hdlib, and experiments note a Cleveland Clinic IBM quantum node with 127 qubits as a potential but unproven accelerator.
AI-written summary · about
chapters (11)
- 0:00Vector-Symbolic Architectures intro
- 1:48Core hyperdimensional ops
- 3:16Encoding data to vectors
- 6:05Retrieving from vectors
- 8:16Currency classification example
- 12:50HD Lib classification model
- 17:45Feature selection with regression
- 20:38Galaxy Platform integration
- 23:47HD Lib implementation
- 25:12Quantum HD computing future
- 26:39Closing & Q&A
related talks
01h 23mVector Databases Demystified: Discover how they work
18mHardware-Aware Static Optimization of Hyperdimensional Computations
50mA fun and absurd introduction to Vector Databases
43mThe Biological Path Towards Strong AI
44mQuantum machine learning: Prospects and challenges
29mDataEngBytes 2023 - SYD-T2-01 - Rene Essomba
54mDissimilarity search...
36mStore and manage data effortlessly with HDF5