Skip to content

Latest commit

 

History

History
34 lines (19 loc) · 1.07 KB

README.md

File metadata and controls

34 lines (19 loc) · 1.07 KB
Raw

GeneSequence

Neon code for Splice Junction Gene Sequence classification (EI / IE / Neither)

Training Splice Junction gene data using Neon Framework

Ref: "https://archive.ics.uci.edu/ml/datasets/Molecular+Biology+%28Splice-junction+Gene+Sequences%29"

Setup

Have neon repo sync'd to local host and follow the steps to build the repo: https://github.com/NervanaSystems/neon

Running "GeneSequence" - Trainig and validation

Place files given in this repo accordingly

gene.py -> ~/neon/neon/data/.

splice_train_val.py -> ~/neon/examples/gene/. (make dir gene)

spliace_data_CAGT.csv -> ~/neon/examples/gene/.

Command to Train and Validate

~/neon/examples/gene/$ python splice_train_val.py

Expected output

(.venv2) username@linuxhost:~/MyProjects/neon/examples/gene$ python splice_train_val.py Epoch 0 [Train |████████████████████| 2552/2552 batches, 0.14 cost, 12.77s] 2017-05-28 22:02:33,525 - neon - DISPLAY - Misclassification error = 0.6%

Known Issues: Temporary fix to avoid evaluation failure due to 'nan'