Differentiable Neural Computers for in silico molecular design: Benchmarks of architectures in generative modeling of molecules

Examensarbete för masterexamen

Please use this identifier to cite or link to this item: https://hdl.handle.net/20.500.12380/300419
Download file(s):
File Description SizeFormat 
CSE 19-94 ODR Prykhodko Johansson.pdfDifferentiable Neural Computers for in silico molecular design: Benchmarks of architectures in generative modeling of molecules4.15 MBAdobe PDFView/Open
Bibliographical item details
FieldValue
Type: Examensarbete för masterexamen
Title: Differentiable Neural Computers for in silico molecular design: Benchmarks of architectures in generative modeling of molecules
Authors: PRYKHODKO, OLEKSII
JOHANSSON, SIMON
Abstract: In the area of in silico drug discovery, deep learning has grown immensely as a field of research. Recurrent neural networks (RNN) is one of the most common approaches used for generative modeling, but recently the differentiable neural computer (DNC) has been shown to give considerable improvement over the RNN for modeling of sequential data. In this thesis, a DNC has been implemented as an extension to REINVENT, an RNN based model that has already been successfully shown to generate molecules with high validity. The model was benchmarked on its capacity to learn the SMILES language on the GDB-13 and ChEMBL datasets. The DNC shows some improvement on all tests conducted at the cost of greatly increased computational time and memory consumption, which puts its practical use into question. This project also gives some insight into the effect of the DNC hyperparameters for the task of generative modeling of molecules.
Keywords: Computer science;machine learning;recurrent neural networks;GRU;LSTM;differentiable neural computer;engineering;project;thesis
Issue Date: 2019
Publisher: Chalmers tekniska högskola / Institutionen för data och informationsteknik
URI: https://hdl.handle.net/20.500.12380/300419
Collection:Examensarbeten för masterexamen // Master Theses



Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.