Luziwei Leng 

Email: luziwei.leng@kip.uni-heidelberg.de

Phone: +49 6221 9847

Office: 01.317, KIP, INF 227

Bio

I am a third year PhD student in the Electronic Vision(s) group from KIP at the University of Heidelberg, where I also obtained my master degree in physics. I developed an interest in cognitive science and artificial intelligence during my bachelor at UESTC in physics and software engineering. Currently, I also worked as a part-time data scientist and consultant for AI related projects.  

CV

Research

Deep learning (DL) related methods have recently achieved great success in solving problems arising from both academy and industry, such as image classification, language translation and Go. However, in terms of the realization of general AI, DL is still far from perfect for many reasons, for example its dependence on huge amount of data and the lack of ability in generalizing what it learned and apply them for new tasks.

One option is to take inspirations from biology. The human brain is organized in a way that high level function areas integrate inputs from multiple sensory paths, which enables it to associate different types of stimuli in cognitive activity. Within this hierarchical structure, local dynamics are characterized with varying temporal scales.

 

Currently, my work mainly focuses on the development of functional networks and corresponding learning algorithms involving biological neuron models and synaptic dynamics, which can solve problems of the same order of complexity as those addressed by traditional artificial neural networks. I am especially interested in how certain biological properties or mechanisms can contribute to improving system functionality. Another goal of my work is to implement the network on neuromorphic platforms, aiming for an improvement in speed, robustness and energy consumption.

At last, I always keep a strong interest in cutting edge machine learning approaches solving challenging problems, such as combination models, reinforcement learning, to name a few. Cases are increasing where computational neuroscience borrow ideas from machine learning.

Publications

Journal

Spiking neurons with short-term synaptic plasticity form superior generative networks Luziwei Leng*, Roman Martel, Oliver Breitwieser, Ilja Bytschok, Walter Senn, Johannes Schemmel, Karlheinz Meier, Mihai A. Petrovici*. Scientific Reports, 2018.07.

Spiking neurons with short-term synaptic plasticity form superior generative networks Luziwei Leng*, Roman Martel, Oliver Breitwieser, Ilja Bytschok, Walter Senn, Johannes Schemmel, Karlheinz Meier, Mihai A. Petrovici*. 2017.09, on arXiv

Conference

Spiking neural networks as superior generative and discriminative models Luziwei Leng*, Mihai A. Petrovici*, Roman Martel, Ilja Bytschok, Oliver Breitwieser, Johannes Bill, Johannes Schemmel, Karlheinz Meier. Cosyne Abstracts 2016, Salt Lake City USA.

Stochastic inference with spiking neural networks Mihai A. Petrovici*, Luziwei Leng*, Oliver Breitwieser*, David Stöckel*, Ilja Bytschok, Roman Martel, Johannes Bill, Johannes Schemmel, Karlheinz Meier. BMC Neuroscience 2016, 17(Suppl 1):P96

Poster

Spiking neural networks as superior generative and discriminative models  Luziwei Leng*, Mihai A. Petrovici*, Roman Martel, Ilja Bytschok, Oliver Breitwieser, Johannes Bill, Johannes Schemmel, Karlheinz Meier. Poster Award, 2016 Neuro-Inspired Computational Elements Workshop (NICE2016), Berkeley, California 

Thesis

Deep Learning Architectures for Neuromorphic Hardware Master's Thesis

 
up