Research
Each line of work starts from a question about the brain. Masters, Honours and PhD students take a piece of it, usually from one of our Requests for Research, and the next student builds on what the last one published.
2 students working now17 alumniMonash · UNSW partner universities
Since 2018 · open Request for Research
Why do brains have two hemispheres, and can the asymmetry improve deep learning, RL and robotics?
3 papers and theses
Working on it now
Before them: Amir, Chandramouli, Yanfeng, Jarrad, Grant, Haowei, Joel, Ben, Jingwen
Since 2021 · open Request for Research
If the hippocampus replays to cortex to build a world model rather than a policy, does that make continual RL more sample-efficient and less forgetful?
2 papers and theses
Working on it now
Before them: Luke
Since 2022 · open Request for Research
Can a world model (frontal cortex) gated by a model-free learner (basal ganglia) explain executive control, and give RL agents the benefits of both?
No papers yet
Working on it now
Open to a new student
Before them: Guixuan, David
Since 2018
How does a fast hippocampal learner beside a slow neocortical one let a system learn specific instances in one shot without forgetting?
2 papers and theses
Working on it now
Open to a new student
Before them: Mohit, Abdelrahman
Since 2026 · open Request for Research
Can a fast episodic memory learn to steer a slow, generalising one, so that few-shot and continual learning work without forgetting?
2 papers and theses
Working on it now
Open to a new student
To keep our work apart from the dominant trajectory of machine learning, we hold ourselves to a set of constraints and themes.
To force a separation between our work and the dominant trajectory of Machine Learning we self-impose "Biological Plausibility" constraints. That doesn't mean we need to do everything exactly as it is done in "Wet Brains": Rather, we aim to exclude the most implausible features of ML and develop biologically-plausible alternatives with similar computational qualities.
Rather than focus on building control and planning systems explicitly, later moving towards more abstract reasoning systems, we take the approach of building a sophisticated simulation of the world first, and then exploiting this simulation for planning and goal-directed behaviour. Our strategy will be to learn to build predictive models, extend predictions many steps into the future in Generative mode, and then learn to control these predictions resulting in the ability to evaluate action choices over many time steps. This is our definition of mental simulation.
In practice, we have found that sparse representations have many desirable properties including combinatoric representational capacity and the ability to represent novel combinations of concepts without additional learning.
Capsules networks learn to describe observations using only a few parameters of a subset of cells. This bottleneck forces Capsule networks to disentangle the latent variables and entities in the world and describe them efficiently. These representations are more powerful and flexible than conventional neural models.
Attetional filtering has recently delivered state of the art results in a number of tasks especially natural language modelling. However, the dominant approach to attentional filtering still relies on deep backpropagation to learn attentional strategies. To avoid biologically implausible time-travel we will use the Bellman equation (a key part of discounted reinforcement learning) to associate delayed rewards with their causes.
Rodney Brooks' early work on the subsumption archhitecture and the importance of embodiment in robotics still resonates with us today. Many of the capabilities and features we intend to demonstrate only fit into an agent-based paradigm, which means we must develop an environment for it to operate in. Fortunately, these days there are a number of excellent (mostly RL-based) simulated environments to play with, such as the OpenAI Gym and the AnimalAI Olympics.
One of the most biologically-implausible aspects of conventional ML is the use of deep-backpropagation as the core mechanism for credit assignment (i.e. learning network parameters). We then face the considerable challenge of building entirely different learning systems that are still able to perform as effectively as those trained with deep-BP.
One of the key constraints we apply is that our methods must learn continually even in nonstationary environments (i.e. the statistics of the data may change at any time, and learning effectiveness must not be impaired by this). Many ML methods fail to function well in nonstationary environments and therefore cannot learn continually. The criterion of "lifelong" learning is also often used, meaning that learning cannot stop.
We assume that external feedback can only take the form of sparse (rare) feedback, whether in the form of sparse labels or rewards. In addition we require that our methods function with only rewards (the quality of a state or action) rather than the ideal state or action. This matches the conditions in which people learn.
Given only sparse global rewards, the vast majority of learning must be either unsupervised or self-supervised. One of the most promising learning rules may be Predictive Coding - learning to predict the next input, whatever it is. The rule needs only local information and naturally learns the dynamics of observed sensor input.
One of the key inspirations for our work was Jeff Hawkins' elimination of the symbol-grounding problem by redefining it as an incremental process of accumulating invariances. Later, researchers generalized this to include notions of equivariance (Hinton) and disentanglement (Y. Bengio). Whichever qualities are produced in the process of abstraction, the transformation must be incremental (and reversible) not discrete.
Having established that representations must be symbolic to varying degrees, and building on the effective functioning of deep neural networks, it is natural that representations should be hierarchical - that is, concepts at different levels in a graph should have varying scope and specificity. For efficiency, the same representation and hierarchy should be used for planning - meaning that plans would not exist as a series of steps but as a set of features in the graph that are incrementally unfolded an developed to match observed reality. This is the process by which we hope to translate mental simulations to effective action in simulated environments.