1,721,123 research outputs found
Processing of analogy in the thalamocortical circuit
No full textThe corticothalamic feedback and the thalamic reticular nucleus have gained
much attention lately because of their integrative and modulatory functions.
A previous study by the author suggested that
this circuitry can process analogies (i.e., the {\em analogy hypothesis}).
In this paper, the proposed model was implemented as a network of leaky
integrate-and-fire neurons to test the {\em analogy hypothesis}.
The previous proposal required specific delay and
temporal dynamics, and the implemented network tuned
accordingly functioned as predicted. Furthermore, these specific
conditions turn out to be consistent with experimental data, suggesting
that a further investigation of the thalamocortical circuit within the {\em
analogical framework} may be worthwhile
A Metric for Intrinsic Motivation in Reinforcement Learning Agents
Full text linkClassically, the reward for an agent is given by extrinsic factors which motivate the agent to improve and learn; however, an active area of research within cognitive science and AI is the effect and necessity of intrinsic motivation for an agent. This can manifest itself in many forms from curiosity to reduction of cognitive dissonance to motivation for effectance. Despite the prevalence and perceived importance of intrinsic motivation, there is no metric to measure ���how��� intrinsically motivated an agent is compared to another and is instead, largely empirical. Furthermore, methods that might be stated to be intrinsically motivated can be directly linked to the environment and thus, might be less intrinsically motivated than thought. Thus this thesis presents a general metric for intrinsically motivated agents to suggest that highly intrinsically motivated agents are more robust than less intrinsically motivated agents. First, an overview and review of reinforcement learning and intrinsic motivation is presented. Following this, a general metric is proposed with empirical and mathematical justification to measure the intrinsic motivation of an agent. Lastly, several intrinsic motivation agents are tested to evaluate the metric and compare the relative performance of the agents
Analysis of Time-Delay Artificial Neural Networks in Ball Catching Task
Full text linkIn this paper, we look at the performance of a time-delay neural network in a scenario requiring memory as well as reactivity. Utilizing a ball catching scenario where the agent will have to move to catch a falling ball, and then remembering where the second one was relative to its position in order to catch the second, we can determine how the time-delay neural networks perform in these tasks. For comparison to previous work with this scenario, we will compare the performance to a feed-forward network and a recurrent neural network
The Effect of Various Temperature Schedules on the Comfort and Energy Efficiency of Radiant Room Heating
Full text linkThe automation of finding a balance between energy consumption and resident comfort has been a core issue with home heating system designers. However, there has been limited research regarding intelligent control in East Asian style radiant-heating home systems, which controls temperature with a heating valve through which hot water runs beneath the floor. While there was one study that attempted to use Reinforcement Learning based off of resident input to control the heating system, our approach is to design a fixed target temperature schedule for the heating valve to adhere to that finds a balance between energy consumption and resident comfort by reducing energy usage as much as possible without an excessive cost to resident comfort. This comes with the assumption that there are typical patterns in the optimal temperature setting for each individual, and that these patterns are relatively constant. We decided to use square-wave policies, which are schedules that have the target temperature alternate between the optimal temperature and 1 ��C lower, and compared between a policy that alternates every hour, another policy that alternates every two hours, and the default policy of always setting the target temperature at the optimal temperature. We tested our square-wave policy in an experimental residential unit, and observed the amount of energy consumed and the average deviation from the optimal temperature setting. Our results show that within a typical 4-hour temperature setting zone, the energy consumption for the square-wave policies was reduced by ~30% ��� ~50% compared to the default policy, while the average temperature deviation of the square wave policies only differentiated from the default by 0.05 ��C, which is too little to have any noticeable effects. However, although we analyzed these determined square-wave policies and determined their benefits, the unique heating and cooling characteristics of a specific residential unit may decide which one to choose or whether further fine-tuning of these schedules is needed. To explore in this direction, we developed an Artificial Neural Network to learn the heating and cooling dynamics in each individual residential unit and predict future temperature fluctuations based on past conditions; the results were very promising with highly accurate predictions. This will allow us to further fine-tune the square-wave interval. We expect these approaches we presented above to enable significant energy savings while maintaining comfortable indoor temperature levels
Second order isomorphism: A reinterpretation and its implications in brain and cognitive sciences
Full text linkShepard and Chipman's second order isomorphism describes how
the brain may represent the relations in the world.
However, a common interpretation of the theory can cause difficulties.
The problem originates from the static nature
of representations. In an alternative interpretation, I propose that
we assign an active role to the internal representations and
relations. It turns out that a collection of such active units can
perform analogical tasks. The new interpretation is supported
by the existence of neural circuits that may be implementing such a function.
Within this framework, perception, cognition, and motor function
can be understood under a unifying principle of analogy
Analyzing Cricket Songs with Machine Learning
Full text linkGiven the recordings of cricket songs, we aimed to apply machine learning and audio signal processing techniques to (1) discover intrageneric and intraspecies relationships in the songs and (2) create models that could classify the songs into their correct genus and species. First, we took out noise in the audio files using a high pass filter. Then, we represented the cricket songs in three different forms: mel spectrograms, mel frequency cepstrum coefficients, and magnitude power spectrums. We achieved our first objective by reducing the dimensionality of the three extracted audio features to visualize how the cricket songs clustered in 2D space. We found that cricket songs belonging to the same genus are generally similar, which gave us hope that constructing a high accuracy genus classification model would be possible. We were not able to conclude that cricket songs belonging to the same species are similar because the dataset used did not have enough audio files per species. As a result of our initial findings, we constructed genus classification models using shallow convolutional neural network architectures and mel spectrograms as input. Because there was only a small number of cricket song files available for training, we reduced our models��� scopes to only classify inputs into 5 genera. Rather than extracting a single mel spectrogram per each available audio file, we extracted multiple 3-second mel spectrograms per each available audio file and used this set of mel spectrograms as our training set. We found that the more mel spectrograms extracted per available audio file, the higher the model���s accuracy became. Our highest-performing genus classification model has a validation loss of 0.1637 and a validation accuracy of 94.30%. Later in the research, we obtained a larger dataset that could be used for species classification. As such, we constructed species classification models using the same approaches taken for the genus classification models. These models classified inputs into 9 different species. Our highest-performing species classification model has a validation loss of 0.2849 and a validation accuracy of 92.28%. The high accuracy of our genus and species classification models confirmed that it would be possible to classify cricket songs using machine learning techniques. This finding is pivotal for the entomology field as there has not been much documented research regarding insect song classification. Of course, our classification models were quite limited in scope; however, we believe that with more data and a deeper model, more generalized insect song classification models are possible. The fact that we were able to employ simple yet effective techniques to discover insights in relatively small cricket song datasets should encourage entomologists that the application of machine learning to cricket songs is still worthwhile even if the field has a lack of publicly available data
Knife-Edge Scanning Microscope Brain Atlas: A Web-Based, Light-Weight 3D Mouse Brain Atlas
No full text
Going Beyond Counting First Authors in Author Co-citation Analysis
Full text linkThe present study examines one of the fundamental aspects of author co-citation analysis (ACA) - the way co-citation
counts are defined. Co-citation counting provides the data on which all subsequent statistical analyses and mappings
are based, and we compare ACA results based on two different types of co-citation counting - the traditional type that
only counts the first one among a cited work's authors on the one hand and a non-traditional type that takes into
account the first 5 authors of a cited work on the other hand. Results indicate that the picture produced through this non-traditional author co-citation counting contains more coherent author groups and is therefore considerably clearer. However, this picture represents fewer specialties in the research field being studied than that produced through the traditional first-author co-citation counting when the same number of top-ranked authors is selected and analyzed. Reasons for these effects are discussed
Automated counting of cell bodies using Nissl stained cross-sectional images
No full textCell count is an important metric in neurological research. The loss in numbers
of certain cells like neurons has been found to accompany not only the deterioration of
important brain functions but disorders like clinical depression as well. Since the manual
counting of cell numbers is a near impossible task considering the sizes and numbers
involved, an automated approach is the obvious alternative to arrive at the cell count. In
this thesis, a software application is described that automatically segments, counts, and
helps visualize the various cell bodies present in a sample mouse brain, by analyzing the
images produced by the Knife-Edge Scanning Microscope (KESM) at the Brain
Networks Laboratory.
The process is described essentially in five stages: Image acquisition, Pre-
Processing, Processing, Analysis and Refinement, and finally Visualization. Nissl
staining is a staining mechanism that is used on the mouse brain sample to highlight the
cell bodies of our interest present in the brain, namely neurons, granule cells and
interneurons. This stained brain sample is embedded in solid plastic and imaged by the
KESM, one section at a time. The volume that is digitized by this process is the data that
is used for the purpose of segmentation.
While most sections of the mouse brain tend to be comprised of sparsely
populated neurons and red blood cells, certain sections near the cerebellum exhibit a
very high density and population of smaller granule cells, which are hard to segment
using simpler image segmentation techniques. The problem of the sparsely populated
regions is tackled using a combination of connected component labeling and template matching, while the watershed algorithm is applied to the regions of very high density.
Finally, the marching cubes algorithm is used to convert the volumetric data to a 3D
polygonal representation.
Barring a few initializations, the process goes ahead with minimal manual
intervention. A graphical user interface is provided to the user to view the processed data
in 2D or 3D. The interface offers the freedom of rotating and zooming in/out of the 3D
model, as well as viewing only cells the user is interested in analyzing. The
segmentation results achieved by our automated process are compared with those
obtained by manual segmentation by an independent expert
- …
