Surprise in the Lab: MIT Scientists Unearth New Aspects of Mouse Intelligence. A recent MIT study on mouse behavior in reward-based tasks showed that mice, while capable of learning the best strategy, often deviate from it, suggesting a complex decision-making process.

MIT study on mouse behavior in reward-based tasks showed that mice, while capable of learning the best strategy, often deviate from it. This finding, using a new analysis tool called blockHMM, has potential implications for neurological research, particularly in understanding conditions like schizophrenia and autism

The finding using a new analysis tool called blockHMM has potential implications for neurological research. The research shows that mice learn the winning strategy in a simple game, but refuse to commit to it. This could help understand conditions like schizophrenia and autism. 

Here are some other findings using blockHMM: 

• Secondary structure information for protein sequences 
• Equivalence allows results from both HMM or local score to be transposed into each other 

BlockHMM is a Block Hidden Markov Model. In the blockHMM generative model, each hidden state governs the choice sequence in each block.

Hidden Markov Models (HMMs) are statistical models that are often used in machine learning. They are used to model data sequences, such as in speech recognition, natural language processing, and bioinformatics. HMMs are probabilistic models that consist of a sequence of hidden states that generate observations. The goal of HMMs is to estimate the sequence of hidden states based on a sequence of observations. 

Here are the steps for the Hidden Markov Model algorithm: 

1. Define the state space and observation space 
2. Define the initial state distribution 
3. Define the state transition probabilities 
4. Define the observation likelihoods 
5. Train the model 
6. Decode the most likely sequence of hidden states 

The forward algorithm is used to calculate a “belief state” in a HMM. This is the probability of a state at a certain time, given the history of evidence

Here are some examples of Hidden Markov Models (HMMs): 

• Weather prediction: Predicting the weather based on the type of clothes someone wears. 
• Stock price forecasting: Using Markov chains to forecast stock prices. 
• Ice cream task: The hidden states are hot and cold weather, and the observations are the number of ice creams eaten on a given day. 

Other examples of HMMs include: 

• How a person feels in different climates 
• A random walk in one dimension 

There are also domain-specific variations of HMMs. For example, in biological sequence analysis, there are profile-HMMs, pair-HMMs, and context-sensitive HMMs

HMM provides solution of three problems : evaluation, decoding and learning to find most likelihood classification. This chapter starts with description of Markov chain sequence labeler and then it follows elaboration of HMM, which is based on Markov chain

In the new open-access study in PLOS Computational Biology, mice surprised scientists by showing that while they were capable of learning the “win-stay, lose-shift” strategy, they nonetheless refused to fully adopt it

A study in PLOS Computational Biology found that mice were able to learn the “win-stay, lose-shift” strategy but refused to fully adopt it. 

A similar reluctance to adopt the “win-stay” strategy has also been observed in human reversal learning and spatial discrimination learning in rats. 

Some possible reasons for this reluctance include: 

• Learning two rules simultaneously 
• Being exploited by competitors 

The “win-stay, lose-shift” strategy can be viewed as too strong for a selfish player. In competitive environments, expressing this strategy can lead to exploitation by competitors

Open access in PLOS Computational Biology means that authors agree to make their articles available for reuse without permission or fees. Anyone can copy, distribute, or reuse these articles, as long as the author and original source are properly cited. 

PLOS was founded in 2001 with the goal of transforming science communication. PLOS journals were some of the first Open Access titles in biology and medicine. 

PLOS Computational Biology also requires authors to make any author generated code underlying the findings fully available without restriction. When submitting a manuscript online, authors must provide a Data Availability Statement describing compliance with both policies. 

The editor-in-chief of PLOS Computational Biology is Feilim Mac Gabhann.

In a new open-access study in PLOS Computational Biology, mice surprised scientists by learning the “win-stay, lose-shift” strategy but not fully adopting it. This task is difficult for people with schizophrenia. 

PLOS Computational Biology is a peer-reviewed journal that focuses on computational studies. The journal was established in 2005 by the nonprofit Public Library of Science. 

Some other topics covered in PLOS Computational Biology include: 

• The neurocognitive role of working memory load 
• Learning context shapes bimanual control strategy 
• Evidence that rodents form an outgroup to humans and dogs 

Mice are often used in neuroscience research because their brains share a lot of structural organization and genetic information with humans. This allows scientists to manipulate the mouse genome to build models of human diseases.

Mice are natural learners and can learn to avoid regret. They can also learn to: 

• Retrieve food Mice can learn to retrieve food from the ends of arms without revisiting depleted ones. 
• Switch strategies Mice can learn to switch from a spatial to a sound-based strategy to retrieve pups. 
• Avoid poisoned baits Rodents can quickly learn to avoid poisoned baits. 
• Navigate Mice can learn to navigate to a water port after just a few reward experiences. 

Mice can learn rapidly, with a phenomenon called “sudden insight”. In traditional mouse learning experiments, such as the steering wheel task, mice tend to learn slowly

Here are some other learning strategies that mice use: 

• Categorize visual stimuli: Mice learn to respond to positive and negative reinforcement in the presence of discriminative stimuli. 
• Use distal cues: Mice can create a spatial map of their environment using distal cues. 
• Switch strategies: Mice can switch between strategies, such as exploring openly and inferring a stable rule. 
• Hunt: Mice can adopt different hunting strategies(full article source google)