Instructor: Michelle R. Greene, Ph.D

Email: mgreene2@bates.edu

Office hours: F 9:30-10:30; Hathorn 106
OR Book me via calendar if this time does not work

Logistics: MWF 8:25am - 9:20 (Carnegie 113)

Prerequisites: NS/PY 160 or 200

Course Description

Although a central tenet of neuroscience is that information about the world in encoded in the patterns of neural firing, it is increasingly acknowledged that our assumptions about these patterns make qualitatively different predictions about neural function. This course examines major hypotheses related to information coding by individual neurons and populations of neurons. Specific themes include rate coding versus time-based codes, sparse versus dense codes, and the relationship between brain responses and the statistics of their inputs. Students examine biological data and artificial models to assess how various encoding schemes might produce skillful behavioral responses.

Introduction

What is a “neural code”? This fundamental concept refers to the rules that transform action potentials into perceptions, concepts, memories, emotions, and actions. The neural code is the software of the brain.

Solving the neural code is the deepest, most foundational problem in neuroscience. Cracking this code will give us unlimited control over our own brains, allowing us to “fix” brains that have been broken through stroke or other injury. It will allow us to truly read minds, let us share our thoughts with others remotely, or even upload our consciousness when we die. More broadly, understanding the code will help us solve long-standing philosophical mysteries such as the mind-body problem and the existence of free will.

Just as solving the genetic code over half a century ago led to an understanding of human variability and disease and has paved the way for editing this code using techniques such as CRISPR-CAS9, solving the neural code will be just as transformative in the next century. In fact, Francis Crick, who co-discovered of the genetic code, spent the second half of his career working on the neural code.

We are still in relatively early days of this endeavor. Currently, we do not have a single neural code, but many. There are rate codes, temporal codes, population codes and grandmother-cell codes, quantum and chaotic and information codes, and codes based on oscillations and synchronies. In this course, we will explore many of these codes with an eye to how specific codes would enable the brain to efficiently solve hard problems.

Learning Objectives:

By the end of this course, you will be able to:

  • Explain the concept of information as a quantifiable entity and be able to compute it mathematically from a probability distribution.

  • Explain the concept of a representation and be able to distinguish between a representation’s content and its function.

  • Identify the constraints that place bounds on the types of neural codes that can exist. Be able to use Fermi estimation to evaluate neural codes in light of these constraints.

  • Compare and contrast competing theories of neural coding, evaluating their comparative strengths and tradeoffs.

Policies and Expectations:

Students with Disabilities or Learning Differences

If you think you need accommodation for a disability or learning difference, please let me know as soon as possible. Some aspects of this course may be modified to facilitate your participation and progress. You will need to create documentation with the Office of the Dean of Students. I will treat any information about your disability with the utmost discretion.

Commitment to Equity and Diversity

Everyone has the right to be treated with dignity and respect in our classroom, and I am committed to making that happen. But this not only means your instructor treating you with respect, but also you treating each other that way. Disrespect or discrimination on any basis will not be tolerated. Whether inside or outside the classroom, if you encounter sexual harassment, sexual violence, or discrimination based on race, color, religion, age, national origin, ancestry, sex, sexual orientation, gender identity/expression, or disability, you are encouraged to report it to Gwen Lexow, Director of Title IX and Civil Rights Compliance at Bates at or 207-786-6445. Additionally, please remember that Bates faculty are concerned about your well-being and development, and we are available to discuss any concerns you have. Students should be aware that faculty are legally obligated to share disclosures of sexual violence, sexual harassment, relationship violence, and stalking with the college’s Title IX Officer to help ensure that your safety and welfare are being addressed.

Academic Integrity

Cheating is bad, I think we can all agree to that. The less acknowledged truth is that it’s not even worth it. Cheating cheapens the value of your work and everyone else’s, and a single violation can literally ruin your entire academic and professional career. Please remind yourself of the Bates College policy on academic integrity and its definitions of plagiarism, use/misuse of sources, and cheating. Students’ work will be closely scrutinized for plagiarism and violations of the College policy will not be tolerated. If you are concerned that your collaboration might put you at risk of an academic integrity violation, please come see me during office hours as soon as possible. In my experience, violations of academic integrity are acts of desperation. If you are ever feeling desperate enough that a few extra points in this course seem to be worth risking so much, please consider talking to someone first — that could be me, a friend, or even someone at CAPS. I want you to succeed, and am happy to talk to you if you are feeling undue pressure from this course or anything else.

Emergencies

If I must cancel class due to weather or an emergency, I will inform you via the class email list. Please consider your Bates email to be the default place to look for class-related information and get into the habit of checking it daily.

Electronics

In our classroom, we are mindful of how we use technology, both for our own learning, and for those around us. You may use a tablet or laptop for taking notes, but I strongly encourage you to avoid having your laptop out for other reasons: laptop use is correlated with lower learning outcomes for you and those around you, and the act of taking notes on the laptop is less effective than hand-written notes.

Emergencies

If I must cancel class due to weather or an emergency, I will inform you via the class email list. Please consider your Bates email to be the default place to look for class-related information and get into the habit of checking it daily.

Because this class takes place in a pandemic…

The ongoing Covid-19 pandemic necessitates flexibility from everyone for this course to take place in a safe and supportive environment. Below are some key considerations:

  • To ensure the safety of other students and in line with Bates College and CDC health policies, everyone must wear a mask in the classroom at all times. Please ensure that your mask covers both your nose and mouth.

  • Some students may need to attend class remotely on a temporary basis due to 1) testing positive for Covid-19, 2) being in close contact with someone who tested positive for Covid-19, or 3) experiencing symptoms of Covid-19 or are otherwise not feeling well. I encourage you to err on the side of caution and attend remotely. If you need to attend class remotely, please watch the live class session after it is posted on Lyceum and email me to schedule an office hours appointment to address any questions you may have about class content.

  • As your instructor, I am working to ensure your safety by being fully vaccinated and boosted. However, if I test positive for Covid-19, have symptoms of Covid-19, or need to quarantine after spending time with someone who tests positive for Covid-19, the class will temporarily be moved to remote format via Zoom.

  • I recognize that this may all seem very unfair. It is. You did not ask for a global crisis, and you may personally have been doing everything possible to do your part. You may be experiencing stress in other ways, such as if someone close to you becomes ill. If you find yourself struggling, please come talk to me. I will help support you and point you towards additional resources.

Zoom Etiquette

  • Please mute your audio when you are not speaking. This will help to limit background noise.
  • Because not all body language is transmitted on video, it’s important to structure who is speaking and when. Please use the “Raise Hand” feature if you would like to speak or answer a question, and please consider letting others know you are finished by saying something like “That’s all”, “I’m done”, or “Thank you”. This will ensure that no one is feeling interrupted.
  • I welcome you to use the chat function for questions and comments, but please be aware that it is public, and a record of the chat is archived.
  • I strongly encourage you to keep your video on. This will help your colleagues and myself feel connected with you. I understand that it’s not always fun to see yourself on camera. If this is the case for you, you may wish to consider hiding your face in your view.
  • Please attend to all personal needs before class (dressing, grooming, eating, etc.). Please also be mindful of your movement during class — a moving video from a mobile phone walking between rooms can be very distracting!

A Note on Email

Email is a good servant, but a bad master. As such, I try to limit my use of it to spend more time on the parts of my job that matter: giving you the best class experience possible, and being a great scientist. Statistically speaking, you probably don’t like email much either. Therefore, I process all of my email in one batch once a day. While I will always respond to you in the same day, this means that you may go 23 hours without a response. Need a faster response? Stop by my office - I’m there most of the time! Need to meet with me? Feel free to send me a calendar invite. Here are step-by-step instructions for this.

Grading

Homework: 40% of total
Starting on the second class period, there will be a set of 1-4 questions about the reading to guide your preparation for class. Please upload your responses to these questions (1-2 sentences each is sufficient) the night before class by midnight on Lyceum. These will be graded on a binary scale (0: no response or fundamentally incorrect response; 1: fundamentally correct response). 50% of assignments will be selected randomly for more feedback, and you are encouraged to meet with me during office hours if you would like more feedback or have questions about a question that were not covered in class.

Exams: 40% of total
The two exams will be given in class and will include a mix of multiple-choice, fill-in, and short answer questions. I will return the graded exam to you within one week. If you are not happy with your grade, you will have the opportunity to make corrections on the exam to earn back half of the lost points. For example, if you earned a 70 on an exam, you will be able to earn up to an 85 by doing corrections.

  • Exam 1: February 7 Will cover foundational materials (codes, information, and representations).

  • Exam 2: March 11 Will cover codes for single neurons and neuronal populations.

Take-home final: 20% of total
The final exam for this course is cumulative and open resource (open note, open reading, open collaboration). It will be released after the last class meeting on April 15, and will be due on Lyceum at the end of the scheduled final exam time: April 19 3:15pm.

Your final percentage score will be assigned a letter grade on the following scale:

Grade Percentage Grade Percentage
A+ >95% B- 71-74%
A 87-94% C+ 67-70%
A- 83-86% C 63-66%
B+ 79-82% D 50-62%
B 75-78% F <50

Reading:

All required papers will be available on this website and Lyceum, and should be done before class.

Course Calendar

Unit 1: What are neural codes?

January 12: Introduction

Goals: Get acquainted with the course; understand the problem of the neural code; practice Fermi estimation.
Reading: No reading expected ahead of time. Review the syllabus after class.
Study questions: None

January 14: What is a neural code?

Goals: Be able to articulate why neural codes may not be obvious; practice Fermi estimation.
Reading: Jerome Lettvin et al “What the Frog’s Eye Tells the Frog’s Brain”.
Study questions:

  • Study question 1: “The connections are such that there is a synaptic path from a rod or cone to a great many ganglion cells, and a ganglion cell receives paths from a great many thousand receptors. Clearly, such an arrangement would not allow for good resolution were the retina meant to map an image in terms of light intensity point by point”. Why is this the case? What alternative arrangement would provide a more photograph-like representation?
  • Study question 2: What is the “code” that seems to be “spoken” by ON, ON-OFF, and OFF cells?
  • Study question 3: The researchers’ strategy was to “present the frog with as wide a range of visible stimuli as we could”. What do you think of this approach? What sort of biases might be introduced into the results?

January 17: MLK DAY: NO CLASS

Goals: Attend the excellent programming throughout the college.
Reading: None
Study questions: None

January 19: What is the content of the environment?

Goals: Refine your understanding of the neural code by enumerating what information in the environment is there to be perceived.
Reading: J.J. Gibson “The Environment as a Source of Stimulation” from The Senses Considered as Perceptual Systems
Study questions:

  • Study question 1: As students of the perceptual systems, Gibson encourages us to focus on a physics that describes physical phenomena at spatial scales relevant to animals (millimeters to kilometers). Provide one example of a situation in which this simplification adds to our understanding of sensory stimulation, and one example where it does not.
  • Study question 2: The chapter notes that as terrestrial animals, we are subject to atmospheric pressure. Why are we not explicitly aware of this constant stimulation? How is this similar to the frog’s visual system that we considered last week?
  • Study question 3: Gibson describes language and visual art as second-order sources of information. What does this mean? What would be the primary source of information as you read these words?

January 21: What is information?

Goals: Be able to define infomation as a quantitative concept. Explain how entropy in information theory related to entropy from physics.
Reading: Read: Pierce “The Origins of Information Theory” from An Introduction to Information Theory: Symbols, Signals, and Noise
Note: Pay attention to the notes in the margins. Some pages are optional.
Study questions:

  • Study question 1: Your reading discusses entropy in both the thermodynamic sense as well as the information/communication sense. What fundamental similarity do these two senses share?
  • Study question 2: “The more we know about what message the source will produce, the less uncertainty, the less the entropy, and the less the information”. Please explain how knowing more about a message means less information.
  • Study question 3: When sending messages via telegraphy, engineers had to deal with noise over the cables and lines. Think about neurons sending messages down axons. What types of noise might exist? What properties of neurons ameliorate some of these issues?

January 24: What is information?

Goals: Become comfortable calculating entropy from some common probability distributions.
Reading: Read: Stone “Information Theory” from Principles of Neural Information Theory: Computational Neuroscience and Metabolic Efficiency
Study questions:

  • Study question 1: Consider Figure 2.3a from your reading, and consider it within the specific context of a random variable, such as the outcome of a coin toss. What does it mean to say that Shannon information is a unit of surprise?
  • Study question 2: What is the entropy of a six-sided die?
  • Study question 3: Let’s say that there are 1800 students at Bates College, and 60 of them are neuroscience majors. What is the entropy of the college, and what is the entropy of neuroscience majors?
  • Study question 4: Which entropy would be highest? Which would be the lowest? Why? (a) Neuroscience majors who are juniors; (b) Students who have taken one chemistry class; (c) Students who have taken an FYS.

January 26: What is information?

Goals: Apply the concepts from information theory to neurons.
Reading: Read: Stone Selections from Chapter 3 from Principles of Neural Information Theory: Computational Neuroscience and Metabolic Efficiency
Study questions:

  • Study question 1: Consider Figure 3.3b from your reading. How many unique binary codes could be created in these eight time bins? How many unique codes could be created from 5 of the 8 bins containing spikes?
  • Study question 2: How much information (in bits) is contained in the previous example?
  • Study question 3: Consider Figure 3.4a and compare it to Figure 2.3a from Monday’s reading. Why doesn’t the coding capacity of a neuron always increase with firing rate?

January 28: What is information?

Goals: Develop strategies for efficiently coding information with neurons.
Reading: Read: Huffman Coding from Introduction to Algorithms by Cormen, Leiserson & Rivest.
Study questions:

  • Study question 1: Consider the final encoding shown in Figure 6 on page 2. What is the binary sequence that represents “a”? What is the binary sequence that represents “e”?
  • Study question 2: Explain how encoding these letters in this fashion compresses the data.

Unit 2: What are representations?

January 31: What is a representation?

Goals: Begin to create a working definition of ‘mental representation’ and articulate why this is not a straightforward task. Reading: Palmer, S. (1978) Fundamental Aspects of Cognitive Representation from Cognition and Categorization.
Study questions:

  • Study question: Palmer writes:“Trying to determine the nature of cognitive representation without first knowing about representation as a general construct is much like trying to determine the nature of oak trees without first knowing about trees as a general class of object.” What is a non-cognitive context that you have heard or used the concept of a representation? Consider the five elements of a representational system from page 262 and map them onto your example as far as is possible.

February 2: What is the difference between a representation’s content and its function?

Goals: Refine our working definition of representations to distinguish between a representation’s content and its function.
Reading: Read: Ballard “On the Function of Visual Representation”
* Recommended reading: Marr (1982) “The Philosophy of the Approach” Especially section 1.2.

Study questions:

  • Study question 1: Ballard notes that most researchers in visual perception take a literalist view of perception. How does Ballard define the literalist view?
  • Study question 2: Name two pieces of evidence that call the literalist view into question.
  • Study question 3: What other evidence have we considered this semester that questions the literalist view?

February 4: Are representations ‘real’?

Goals: Explain the problem of the homunculus and question what is representation is for.
Reading:

Study questions:

  • Study question 1: Explain the problems of the homunculus / Cartesian theatre.
  • Study question 2: By contrast, what is the “multiple drafts” theory of consciousness? How does it characterize the observer’s “stream of consciousness”?
  • Study question 3: Imagine that you are reading a journal article about a neural decoding study that shows that information about entity X can be read out from brain area Y (e.g. face identity information in the fusiform gyrus). Is it reasonable to infer that brain area Y is recognizing entity X? Why or why not?

February 7: Midterm 1

No class will be held to give you time on the exam

Unit 3: What are the major types of neural codes?

February 9: What does a single neuron know?

Goals: Explain the concept of a ‘grandmother cell’ and its history
Reading: Read: Gross (2002) Geneology of the “grandmother cell”
Study questions:

  • Study question 1: What is a “grandmother cell”? What do these cells respond to?
  • Study question 2: What is the labeled line theory of neural function? How does it relate to the theory of “grandmother cells”?
  • Study question 3: If you were designing a brain, would a grandmother cell code seem like a good idea to you? Name one or more advantages to such a scheme and one or more disadvantages as you see them.

February 11: What does a single neuron know?

Goals: Determine whether we have evidence for grandmother cell-like coding
Reading: Read: Quiroga et al (2005) Invariant visual representation by single neurons in the human brain.
Study questions:

  • Study question 1: Briefly describe the methodology of these experiments: who were the observers? Where in the brain were the recordings made? What stimuli were the observers shown?
  • Study question 2: Consider the “Jennifer Aniston” neuron presented in Figure 1 and the “Halle Barry” neuron presented in Figure 2. Which neuron seems more like a grandmother cell? Why?

February 14: What does a single neuron know?

Goals: Assess the plausibility of grandmother cell coding
Reading: Read: Excerpts from Anderson An Introduction to Neural Networks, chapter 10.
Study questions:

  • Study question 1: List four issues with grandmother cell representations. Which seems the most serious to you? Which seems the least serious? Why?
  • Study question 2: Your reading describes similarity as a “tricky” concept. Describe in detail one example where “similarity” is not a singular or straightforward concept.
  • Study question 3: For each of the five desirable properties of data representations, provide an example of this in the brain. For example, under the first property (similar events should give rise to similar representations), you might point to the fact that pattern classification algorithms are able to predict what one was seeing, thinking, or doing based on brain data. This means observed neural patterns generalize to new examples.
  • Class activity: here is a link that shows you 20 random articles from Wikipedia

February 16: What does a population of neurons know?

Goals: Compare and contrast a local grandmother cell code with an alternative: distributed codes
Reading: Read: Thorpe (1989) Local versus distributed coding Intellectica
Study questions:

  • Study question 1: Explain the difference between value coding and variable coding, as attributed to Ballard.
  • Study question 2: Is the labeled line hypothesis a local or distributed code? Why?
  • Study question 3: One of the arguments against local coding is that there are too many neurons that would be required. Explain why this line of reasoning is fallacious.
  • Study question 4: Local coding schemes are often criticized for being fragile, meaning that any loss of a neuron could lead to catastrophic results. Thorpe shows that distributed coding schemes might be even worse. Explain this argument.

February 18: What does a population of neurons know?

Goals: Assess the possible advantages of a distributed coding scheme
Reading: Read: Hinton, McClelland, and Rumelhart Chapter 1, Parallel Distributed Processing
Study questions:

  • Study question 1: Figure 2 demonstrates some of the contextual influences on reading. Imagine that the letters with ink splotches were entirely missing - would the letters be similarly easy to guess? What information are you using to perform the task?
  • Study question 2: The chapter discusses frames, scripts, or schemata (synonymous terms for our purposes). What is in your script for a college course?
  • Study question 3: Define what a pattern associator is, and show an example in matrix form.

February 21: WINTER RECESS

February 23: WINTER RECESS

February 25: WINTER RECESS

February 28: What does a population of neurons know?

Goals: Dissect an example of a population code.
Reading: Read: Georgeopoulos (1986) Neural Coding of Movement Direction Science
Study questions:

  • Study question 1: Explain the finding depicted in Figure 1. In what other sensory system have we previously seen a similar type of tuning?
  • Study question 2: Explain the finding depicted in Figure 3. How was the population vector obtained?
  • Study question 3: Are there any biases in the data collection methods that could have influenced this result? What are they?

March 2: How does the brain represent information in time?

Goals: Understand how time provides challenges and opportunities in the study of neural codes.
Reading: Read: Gerster “How can brains be so fast?” from 23 Problems in Systems Neuroscience
Study questions:

  • Study question 1: Explain how the speed of neural processing is limited by the the dynamics of membrane responses and synaptic response. What causes these responses to lag in time?
  • Study question 2: Explain why a “naive” rate coding scheme with a 100 ms integration window would only allow the brain to process at slower than 10 Hz.
  • Study question 3: “Given the large number of neurons in the cortex, another potential coding scheme seems to be a ‘rate’ defined by a population average rather than a temporal average”. If this is the case, would this be a local code, a distributed code, or could it be either? Explain your reasoning.
  • Study question 4: Explain how having a small amount of baseline firing can effectively reduce the membrane time constant.
  • Study question 5 (optional): What questions do you have about the reading that you would like to cover in class?

March 4: How does the brain represent information in time?

Goals: Be able to explain one temporal code: the time to first spike code.
Reading: Read: Van Rullen et al (2005) Spike times make sense. Trends in Neurosciences
Study questions:

  • Study question 1: Describe the experiment cited by the authors on page 1 that provides evidence for information being carried in a “time-to-first-spike” manner. What is the evidence that this code is reliable?
  • Study question 2: Time-based codes are often criticized for being too unreliable due to the natural variability of neural firing. To address this issue, Van Rullen writes “a spike that would be precisely timed with respect to an internal event to which the experimenter does not have access will be considered, by default, as unreliable.” List two types of internal events that he might be referring to.
  • Study question 3: One other issue with time-based codes is that they seem to require a mechanism to keep track of event times and to reset after previous events. What two mechanisms does Van Rullen suggest?
  • Study question 4 (optional): What questions do you have about the reading that you would like to cover in class?

March 7: How does the brain represent information in time?

Goals: Understand how information may be coded in population oscillations.
Reading: Read: Lisman & Jensen (2013) The Theta-Gamma Neural Code. Neuron
Study questions:

  • Study question 1: Consider the theoretical framework presented in Figure 1b. Where is the theta signal represented? Where is the gamma signal represented? How are two unique items represented in this theta-gamma coupling?
  • Study question 2: Is the code presented in this reading a local or distributed code? Why?
  • Study question 3: The authors note that the typical working memory capacity of 7 +/- 2 is the same as the number of gamma cycles within a theta cycle. How compelled are you by the argument that this is a neural correlate of working memory capacity?
  • Study question 4 (optional): What questions do you have about the reading that you would like to cover in class?

March 9: Is spike time variability signal or noise?

Goals: Assess the potential reliability of time-based codes.
Reading: Read: Stein et al (2005) Neuronal Variability: Noise or part of the signal? Nature Reviews Neuroscience Study questions:

  • Study question 1: Explain Figure 1 in your own words. To what extent is the problem illustrated in this figure a problem for rate codes? To what extent is this problem an issue for time-based codes?
  • Study question 2: Explain how noise might be beneficial to the neural code.
  • Study question 3 (optional): What questions do you have about the reading that you would like to cover in class?

March 11: Midterm 2

No class. Exam will be released on Lyceum at 8:25.

Unit 4: What constraints are places on neural codes?

March 14: Why do we have brains?

Goal: Develop a theory of what problems a brain is trying to solve.
Reading: Read: Sterling & Laughlin “Why an Animal Needs a Brain” from Principles of Neural Design
Study questions:

  • Study question 1: Consider the “behaviors” evident in a single-celled bacterium such as E. coli. How is it able to perceive, move, and remember without a brain? How are memories stored in the organism?
  • Study question 2: Why is a longer memory of no advantage to a single-celled organism such as E. coli or Paramecium?
  • Study question 3: Why are fewer synapses less reliable in general? What trade-off does C Elegans make to “get away with” so few synapses?
  • Study question 4: Why is it an advantage for an organism to minimize “wiring” length - i.e. the length of axons?
  • (Optional) Does C. Elegans experience pleasure?
  • (Optional) What other questions do you have about the reading?

March 16: What are the metabolic demands of the brain?

Goal: Begin to build an energy budget for the brain.
Reading: Read: Attwell & Laughlin (2001) An Energy Budget for Signaling in the Grey Matter of the Brain Journal of Cerebral Blood Flow and Metabolism (The calculations on page 1134 are an optional quantitative adventure)
Study questions:

  • Study question 1: If 200,000 Na+ enter the post-synaptic neuron, why are 67,000 ATP required to evict them via Na+/K+ pumps?
  • Study question 2 Pay particular attention to the section “Distributed coding, energy use, and coding sparseness”. Although we will cover the calculations in detail in the class, have a sense of whether local or distributed codes are more energy efficient and why.
  • Study question 3: What are some of the simplifying assumptions made by this work? Will they systematically overestimate or underestimate energy consumption?
  • (Optional): if 3x10^9 ATP are used per neuron, per second; and if 1 mole of ATP contains 30.5 kJ or 7.5 kcal of energy, how many kcal of food is necessary just to power your brain for a day?
  • (Optional): what other questions do you have about this reading?

March 18: How do the brain’s metabolic demands constrain the neural code?

Goal: Be able to use the metabolic demands of the brain as a constaint on the types of codes the brain might use.
Reading: Read: Lennie (2003) The Cost of Cortical Computation Current Biology
Study questions:

  • Study question 1: Name one known difference between human and rodent brains that makes Lennie’s calculations of neural energy use different from those of Attwell & Laughlin from Wednesday’s reading.
  • Study question 2: Describe the methodology that Lennie used to estimate the sustainable spike rate.
  • Study question 3: Name one fundamental agreement between the Lennie’s analysis and that of Attwell & Laughlin.
  • (Optional) what other questions do you have about this reading?

March 21: How redundant is the environment?

Goal: Articulate how we might begin to measure the amount of information in our environments.
Reading: * Read: Attneave (1954) Some Informational Aspects of Visual Perception Psychological Review
Study questions:

  • Study question 1: What does Attneave mean when he considers the visual world to be redundant?
  • Study question 2: Use the Hartley equation to verify the maximum entropy that Attneave calculated in footnote 4.
  • Study question 3: Reflect on the following paradox: how is it that a redundant figure, such as that in Figure 1 as well as a non-redundant figure, such as that in Figure 4, both be perceived as homogeneous?
  • (Optional) What other questions do you have about this reading?

March 23: WINTER RECESS

March 25: WINTER RECESS

March 28: How redundant is the environment?

Goals: Conduct a case study on one method for measuring the information in an environment.
Reading: Read: Kersten (1987) Predictability and redundancy of natural images

Study questions:

  • Study question 1: Kersten uses images that were 128x128 pixels in size, and could take one of 16 different gray level values. How many possible images could be created with these parameters?
  • Study question 2: Kersten (and Shannon in the optional reading) use human judgments to measure entropy. Why is this necessary?
  • Study question 3: How is Kersten’s experiment similar to the thought experiment presented in the first two pages of Attneave’s study that you read last time? How is it different?
  • (Optional) What other questions do you have about this reading?

Unit 5: What are the open questions in neural codes?

March 30: What neural representations are the most compact?

Goal: Be able to articulate how we can leverage the redundancy of the environment to create efficient neural codes.
Reading: * Read: Barlow (1961) Possible Principles Underlying the Transformations of Sensory Messages
Study questions:

  • Study question 1: Describe the “password hypothesis” put forth by Barlow. Describe one feature or entity that you would test as a possible “password” for human observers. (This second part cannot be found in the paper and is for your reflection).
  • Study question 2: Barlow articulates the redundancy reduction hypothesis is “for a given class of input message, it will choose the code that requires the smallest average expenditure of impulses in the output. Or putting it briefly, it economizes impulses”. Imagine that sensory systems are using a Huffman-like code. In what way are energy expenditures being minimized?
  • Study question 3: Are the least redundant parts of a sensory message always the most important for our survival? Why or why not?
  • (Optional) What other questions do you have about this reading?

April 1: Should the brain make the most compact code?

Goal: Make an informed assessment about whether the goal of the neural code is to be maximally efficient.
Reading: Read: Barlow (2001) Redundancy reduction revisited. Network: Computational Neural Systems
Study questions:

  • Study question: Horace Barlow from 1961 gets in a time machine to visit Horace Barlow from 2001. What is one thing that they agree on? What is one thing they disagree on?
  • (Optional): What other questions do you have about this reading?

April 4: Are there silent neurons?

Goal: Assess the extent to which we fully know sensory areas
Reading: Olshausen & Field “What’s the other 85% of V1 doing?” From 23 Problems in Systems Neuroscience
Study questions:

  • Study note: the article frequency references Gabor functions - these are the product of a sine wave and a Gaussian.
  • Study question 1: Describe the biases that we face in neuron sampling? What causes these biases?
  • Study question 2: Describe the biases that we have from the stimuli that are shown to experimental subjects. What might be a way of ameliorating these issues?
  • Study question 3: Describe what is meant by a “classical receptive field”. Describe a result that challenges this notion.

April 6: What is the goal of sensory systems?

Goal: Define the notion of a sparse code and compare it to compact codes.
Reading: * Read: Field (1994) What is the goal of sensory coding?. Sections 1, 2, 5, and 6 required. Others optional, but recommended.
Study questions:

  • Study question 1: What is the difference between a compact code and a sparse code?
  • Study question 2: Explain the concept of a state-space.
  • Study question 3: Be able to explain the transformation that takes place in Figure 2. In what way is the new representation more efficient?
  • (Optional) What other questions do you have about this reading?

April 8: What is a sparse code?

Goal: Develop a working definition of a sparse codes and explain its benefits.
Reading: Olshausen & Field (2004) Sparse Coding of Sensory Inputs Current Opinion in Neurobiology
Study questions:

  • Study question 1: What is meant by the authors when they write about an “overcomplete” representation?
  • Study question 2: Explain the difference between ‘lifetime sparseness’ and ‘population sparseness’. Why is it the case that one does not necessarily imply the other?
  • (Optional) What other questions do you have about this reading?

April 11: How general is the neural code?

Goal: Consider the extent to which the neural code is the same across sensory modality and species.
Reading: McAlpine and Palmer How general are neural codes in sensory systems? from 23 Problems in Systems Neuroscience
Study questions:

  • Study question 1: Briefly describe the differences between barn owls and mammals for coding binaural sound information.
  • Study question 2: How do differences in an animal’s body give rise to differences in the neural code?
  • (Optional) What other questions do you have about this reading?

April 13: Does the neural code change over time?

Goal: Gaze out on the frontier of knowledge and consider what we still don’t understand.
Reading: Yong (2021) Neuroscientists have discovered a phenomenon they can’t explain The Atlantic
Study questions:

Study question 1: Briefly define the concept of representational drift and provide an example of it from the reading.
Study question 2: In what way(s) is representational drift a problem for the way that we think about neural codes? In what way(s) might it be beneficial?
* (Optional) What other questions do you have about this reading?

April 15: What questions do you still have? (Wrap up, reflection, and review)

Goal: Wrap up, reflect, review for final exam.
Reading: None
Study guide: Look at the practice final that is posted on Lyceum and come with your questions.