Q n A with Dr Denis Bauer
summary
Genomics generates a three billion-letter reference and is on track for 50% sequenced by 2025, producing roughly 20 exabytes per year, so traditional single-node compute and naive feature matrices fail — a full three billion-feature matrix is infeasible for current deep nets. Work uses distributed frameworks like Hadoop and Spark to build disposable, question-specific pipelines rather than hardware-locked accelerators, while deep neural networks are being applied successfully to local problems such as predicting deleterious mutations using sequence, epigenome and folding context.
AI-written summary · about
chapters (7)
- 0:00Introduction
- 1:28Key talking points
- 2:21How much has the hype run
- 3:23What about hardware
- 4:22Open source deep learning tools
- 5:48Machine learning and data science
- 6:28How are you finding ODSC India
related talks
28mSome Current Computational Challenges in Bioinformatics
25mScaling Up Genomics with Spark
14mBig data, AI, the genome, and everything
42mHow Novel Compute Technology Transforms Life Science Research
50mCancer Genomics - a biologist and a developer
23mGenomics and AI on Cloud Foundry
21mInteractive Machine Learning on Genomics Data with Spark Notebook
38mInterpreting Deep Learning Models Of Functional Genomics Data To Decode Regulatory Sequence...