start
Differences
This shows you the differences between two versions of the page.
Both sides previous revisionPrevious revisionNext revision | Previous revisionLast revisionBoth sides next revision | ||
start [2015/10/01 14:36] – hj | start [2016/01/22 01:13] – hj | ||
---|---|---|---|
Line 1: | Line 1: | ||
====== iFLYTEK Laboratory for Neural Computing for Machine Learning (iNCML) ====== | ====== iFLYTEK Laboratory for Neural Computing for Machine Learning (iNCML) ====== | ||
- | Welcome to the homepage of iFLYTEK Laboratory for Neural Computing | + | Welcome to the homepage of **iFLYTEK Laboratory for Neural Computing |
{{: | {{: | ||
Line 8: | Line 8: | ||
\\ | \\ | ||
**iFLYTEK Laboratory for Neural Computing \\ | **iFLYTEK Laboratory for Neural Computing \\ | ||
- | for Machine Learning (iNCML)** \\ | + | and Machine Learning (iNCML)** \\ |
Lassonde 2054, Department of Electrical \\ | Lassonde 2054, Department of Electrical \\ | ||
Engineering and Computer Science \\ | Engineering and Computer Science \\ | ||
Line 19: | Line 19: | ||
===== Research Aims of the Lab ===== | ===== Research Aims of the Lab ===== | ||
- | - **Explore new neural computing models for machine learning:** we explore | + | - **Explore new neural computing models for machine learning:** we explore |
- | computing models and algorithms to take advantage | + | - **Investigate neural representations of knowledge for artificial cognition: |
- | - **Investigate neural representations of knowledge for artificial cognition: | + | - **Advance machine intelligence in speech recognition and understanding, |
- | - **Advance machine intelligence in speech recognition and understanding, | + | |
- | |||
- | - **Explore new neural computing models for machine learning:** | ||
- | With the help of advanced computing resources (particularly the general purpose GPU computing platform), we will explore new neural | ||
- | computing models and algorithms to take advantage of big data available in today’s mobile Internet era. Particularly, | ||
- | novel effective unsupervised learning algorithms for neural networks to explore abundant real world unlabeled data for self-learning and | ||
- | adaptation. Moreover, we will research more advanced neural models with long and/or short memory capabilities to explore sequential | ||
- | information within a longer context window for more complex AI tasks, such as human-machine dialogues. | ||
- | |||
- | - **Investigate neural representations of knowledge for artificial cognition** | ||
- | Neural models have achieved huge successes in data modeling. Comparing with data modeling, it is a more challenging problem to represent | ||
- | world knowledge. Knowledge representation requires organizing human knowledge (including common sense, common knowledge and | ||
- | domain-specific information) in an orderly way, as opposed to learning statistical models from all pooled data sets in data modeling. We will | ||
- | investigate a new research direction, named as neural representation of knowledge, to represent all relevant concepts, discrete in nature, as | ||
- | distributed representations in continuous semantic spaces and use a large-scale artificial neural networks (named as semantic brain) to store all | ||
- | possible relations and sematic links among these concepts. The advantages of this new approach lie in two aspects: i) Distributed | ||
- | representations of discrete concepts allow to use regular learning methods for knowledge presentation and the networks may be expanded in | ||
- | size to store as many conceptual relations as needed; ii) All concepts and their relations are stored as distributed representations in semantic | ||
- | brain, allowing to use some heuristic search strategies to perform the basic thinking in a humanoid way, such as reasoning and association. | ||
- | |||
- | - | ||
- | |||
- | |||
- | The new neural computing models and algorithms will be applied to some multimedia AI tasks involving speech, language and image/ | ||
- | More particularly, | ||
- | applications include personal assistant agent running in smart phones. ii) Deep natural language processing and understanding. Its typical | ||
- | applications include automatic machine Q&A systems in general domain, like semantic search engines, or some particular query system in | ||
- | some domains, like medical, health, legal and so on. iii) Image and video scene analysis. Its typical applications include autonomous robot | ||
- | navigation and controlling. | ||
- | |||
- | |||
- | |||
- | ===== Address ===== | ||
- | |||
- | **iFLYTEK Laboratory for Neural Computing for Machine Learning (iNCML)** \\ | ||
- | Lassonde 2054, Department of Electrical Engineering and Computer Science \\ | ||
- | York University, 4700 Keele Street, Toronto, Ontario |
start.txt · Last modified: 2021/05/28 19:05 by hj